Data Structures
class	accumulatable
	abstract class that knows how to add More...

class	algstrct
	algstrct (algebraic structure) defines the elementwise operations computed in each tensor contraction, virtual classes defined in derived typed classes or algstrctcpy More...

class	Bifun_Term

class	bivar_function
	untyped internal class for triply-typed bivariate function More...

struct	BoolPair

class	CommData

struct	CompPair

struct	CompPtrPair

class	ConstPairIterator

class	Contract_Term
	An experession representing a contraction of a set of tensors contained in operands. More...

class	contraction
	class for execution distributed contraction of tensors More...

class	COO_Matrix
	serialized matrix in coordinate format, meaning three arrays of dimension nnz are stored, one of values, and two of row and column indices More...

class	CSR_Matrix
	abstraction for a serialized sparse matrix stored in column-sparse-row (CSR) layout More...

class	ctr

class	ctr_2d_general

class	ctr_replicate

class	ctr_virt

class	CubicModel
	Cubic performance models, which given measurements, provides new model guess. More...

class	distribution

class	endomorphism
	untyped internal class for singly-typed single variable function (Endomorphism) More...

class	grid_wrapper

struct	int1

struct	int2

struct	IntPair

struct	iparam

class	LinModel
	Linear performance models, which given measurements, provides new model guess. More...

class	mapping

struct	mem_loc

struct	mem_transfer

class	Model

class	offload_arr
	offloaded array/buffer More...

class	offload_tsr
	offloaded and serialized tensor data More...

class	PairIterator

class	scaling
	class for execution distributed scaling of a tensor More...

class	scl

class	scl_virt

class	seq_tsr_ctr

class	seq_tsr_scl

class	seq_tsr_spctr

class	seq_tsr_spsum

class	seq_tsr_sum

struct	ShortPair

class	spctr

class	spctr_2d_general

class	spctr_pin_keys

class	spctr_replicate

class	spctr_virt

class	strp_ctr

class	strp_scl

class	strp_sum

class	strp_tsr

class	Sum_Term

class	summation
	class for execution distributed summation of tensors More...

class	tensor
	internal distributed tensor class More...

struct	tensor_name_less
	comparison function for sets of tensor pointers This ensures the set iteration order is consistent across nodes More...

class	Term
	a term is an abstract object representing some expression of tensors More...

struct	time_param

class	topology

class	tspsum

class	tspsum_map

class	tspsum_permute

class	tspsum_pin_keys

class	tspsum_replicate
	performs replication along a dimension, generates 2.5D algs More...

class	tspsum_virt

class	tsum

class	tsum_replicate
	performs replication along a dimension, generates 2.5D algs More...

class	tsum_virt

class	Unifun_Term

class	univar_function
	untyped internal class for doubly-typed univariate function More...

Typedefs
typedef bool	TYPE1

typedef int	TYPE2

typedef int64_t	TYPE3

typedef float	TYPE4

typedef double	TYPE5

typedef std::complex< float >	TYPE6

typedef std::complex< double >	TYPE7

typedef int16_t	TYPE8

typedef int8_t	TYPE9

Enumerations
enum	{ SUCCESS, ERROR, NEGATIVE }

enum	map_type { NOT_MAPPED, PHYSICAL_MAP, VIRTUAL_MAP }

enum	TOPOLOGY { TOPOLOGY_GENERIC, TOPOLOGY_BGP, TOPOLOGY_BGQ, TOPOLOGY_8D, NO_TOPOLOGY }

Functions
void	update_all_models (MPI_Comm comm)

void	factorize (int n, int nfactor, int *factor)
	computes the size of a tensor in packed symmetric layout More...

template<typename ptype >
void	get_perm (int perm_order, ptype A, ptype B, ptype C, ptype &tA, ptype &tB, ptype &tC)

void	calc_fold_lnmk (tensor const A, tensor const B, tensor const C, int const idx_A, int const idx_B, int const idx_C, int const ordering_A, int const ordering_B, iparam *inner_prm)
	calculate the dimensions of the matrix the contraction gets reduced to (A, B, and C may be permuted) More...

void	get_len_ordering (tensor const A, tensor const B, tensor const C, int const idx_A, int const idx_B, int const idx_C, int new_ordering_A, int new_ordering_B, int **new_ordering_C)
	find ordering of indices of tensor to reduce to DGEMM (A, B, and C may be permuted More...

int	ctr_2d_gen_build (int is_used, CommData global_comm, int i, int virt_dim, int &cg_edge_len, int &total_iter, tensor A, int i_A, CommData &cg_cdt_A, int64_t &cg_ctr_lda_A, int64_t &cg_ctr_sub_lda_A, bool &cg_move_A, int blk_len_A, int64_t &blk_sz_A, int const virt_blk_len_A, int &load_phase_A, tensor B, int i_B, CommData &cg_cdt_B, int64_t &cg_ctr_lda_B, int64_t &cg_ctr_sub_lda_B, bool &cg_move_B, int blk_len_B, int64_t &blk_sz_B, int const virt_blk_len_B, int &load_phase_B, tensor C, int i_C, CommData &cg_cdt_C, int64_t &cg_ctr_lda_C, int64_t &cg_ctr_sub_lda_C, bool &cg_move_C, int blk_len_C, int64_t &blk_sz_C, int const *virt_blk_len_C, int &load_phase_C)
	sets up a ctr_2d_general (2D SUMMA) level where A is not communicated function will be called with A/B/C permuted depending on desired alg More...

void	inv_idx (int order_A, int const idx_A, int order_B, int const idx_B, int order_C, int const idx_C, int order_tot, int **idx_arr)
	invert index map More...

template<int idim>
void	spA_dnB_dnC_ctrloop (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, int order_A, int const edge_len_A, int64_t const lda_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

template<>
void	spA_dnB_dnC_ctrloop< 0 > (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, int order_A, int const edge_len_A, int64_t const lda_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

template<>
void	spA_dnB_dnC_ctrloop< MAX_ORD > (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, int order_A, int const edge_len_A, int64_t const lda_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

void	spA_dnB_dnC_seq_ctr (char const alpha, char const A, int64_t size_A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, bivar_function const func)

char *	bcast_step (int edge_len, char A, bool is_sparse_A, bool move_A, algstrct const sr_A, int64_t b_A, int64_t s_A, char buf_A, CommData cdt_A, int64_t ctr_sub_lda_A, int64_t ctr_lda_A, int nblk_A, int64_t const size_blk_A, int &new_nblk_A, int64_t &new_size_blk_A, int64_t *offsets_A, int ib)

char *	reduce_step_pre (int edge_len, char C, bool is_sparse_C, bool move_C, algstrct const sr_C, int64_t b_C, int64_t s_C, char buf_C, CommData cdt_C, int64_t ctr_sub_lda_C, int64_t ctr_lda_C, int nblk_C, int64_t const size_blk_C, int &new_nblk_C, int64_t &new_size_blk_C, int64_t offsets_C, int ib, char const &rec_beta)

void	reduce_step_post (int edge_len, char C, bool is_sparse_C, bool move_C, algstrct const sr_C, int64_t b_C, int64_t s_C, char buf_C, CommData cdt_C, int64_t ctr_sub_lda_C, int64_t ctr_lda_C, int nblk_C, int64_t size_blk_C, int &new_nblk_C, int64_t &new_size_blk_C, int64_t offsets_C, int ib, char const &rec_beta, char const beta, char &up_C, char &new_C, int n_new_C_grps, int &i_new_C_grp, char *new_C_grps)

int	spctr_2d_gen_build (int is_used, CommData global_comm, int i, int virt_dim, int &cg_edge_len, int &total_iter, tensor A, int i_A, CommData &cg_cdt_A, int64_t &cg_spctr_lda_A, int64_t &cg_spctr_sub_lda_A, bool &cg_move_A, int blk_len_A, int64_t &blk_sz_A, int const virt_blk_len_A, int &load_phase_A, tensor B, int i_B, CommData &cg_cdt_B, int64_t &cg_spctr_lda_B, int64_t &cg_spctr_sub_lda_B, bool &cg_move_B, int blk_len_B, int64_t &blk_sz_B, int const virt_blk_len_B, int &load_phase_B, tensor C, int i_C, CommData &cg_cdt_C, int64_t &cg_spctr_lda_C, int64_t &cg_spctr_sub_lda_C, bool &cg_move_C, int blk_len_C, int64_t &blk_sz_C, int const *virt_blk_len_C, int &load_phase_C)

template<int idim>
void	sym_seq_ctr_loop (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

template<>
void	sym_seq_ctr_loop< 0 > (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

template void	sym_seq_ctr_loop< MAX_ORD > (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, uint64_t const offsets_C, bivar_function const func, int const idx, int const rev_idx_map, int idx_max)

void	compute_syoff (int r, int len, algstrct const sr, int const edge_len, int const sym, uint64_t offsets)

void	compute_syoffs (algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, int tot_order, int const rev_idx_map, uint64_t &offsets_A, uint64_t &offsets_B, uint64_t *&offsets_C)

int	sym_seq_ctr_ref (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const *idx_map_C)
	performs symmetric contraction with reference (unblocked) kernel More...

int	sym_seq_ctr_cust (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, bivar_function const func)
	performs symmetric contraction with custom elementwise function More...

int	sym_seq_ctr_inr (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, char const beta, char C, algstrct const sr_C, int order_C, int const edge_len_C, int const sym_C, int const idx_map_C, iparam const prm, bivar_function const *func)
	performs symmetric contraction with blocked gemm More...

void	init_rng (int rank)
	initialized random number generator More...

double	get_rand48 ()
	returns new random number in [0,1) More...

template<typename type >
int	conv_idx (int order, type const cidx, int *iidx)

template<typename type >
int	conv_idx (int order_A, type const cidx_A, int iidx_A, int order_B, type const cidx_B, int **iidx_B)

template<typename type >
int	conv_idx (int order_A, type const cidx_A, int iidx_A, int order_B, type const cidx_B, int *iidx_B, int order_C, type const cidx_C, int **iidx_C)

template int	conv_idx< int > (int, int const , int *)

template int	conv_idx< char > (int, char const , int *)

template int	conv_idx< int > (int, int const , int , int, int const , int **)

template int	conv_idx< char > (int, char const , int , int, char const , int **)

template int	conv_idx< int > (int, int const , int , int, int const , int *, int, int const , int **)

template int	conv_idx< char > (int, char const , int , int, char const , int *, int, char const , int **)

void	flops_add (int64_t n)

int64_t	get_flops ()

void	handler ()

void	cvrt_idx (int order, int const lens, int64_t idx, int idx_arr)

void	cvrt_idx (int order, int const lens, int64_t idx, int *idx_arr)

void	cvrt_idx (int order, int const lens, int const idx_arr, int64_t *idx)

bool	get_mpi_dt (int64_t count, int64_t datum_size, MPI_Datatype &dt)
	gives a datatype for arbitrary datum_size, errors if exceeding 32-bits More...

int64_t	sy_packed_size (int order, const int len, const int sym)
	computes the size of a tensor in SY (NOT HOLLOW) packed symmetric layout More...

int64_t	packed_size (int order, const int len, const int sym)
	computes the size of a tensor in packed symmetric (SY, SH, or AS) layout More...

int	alloc_ptr (int64_t const len_, void **const ptr)
	alloc abstraction More...

int	mst_alloc_ptr (int64_t const len, void **const ptr)
	mst_alloc abstraction More...

void *	alloc (int64_t const len)
	alloc abstraction More...

void *	mst_alloc (int64_t const len)
	mst_alloc allocates buffer on the specialized memory stack More...

int	cdealloc (void *ptr)
	free abstraction More...

template<typename dtype >
const char *	get_fmt ()
	return format string for templated type More...

template<>
const char *	get_fmt< float > ()

template<>
const char *	get_fmt< double > ()

template<>
const char *	get_fmt< int > ()

template<>
const char *	get_fmt< int64_t > ()

template<typename dtype >
void	parse_sparse_tensor_data (char *lvals, int order, dtype const pmulid, int lens, int64_t nvals, CTF::Pair< dtype > pairs, bool with_vals)
	parse string containing sparse tensor into data More...

template<typename dtype >
char *	serialize_sparse_tensor_data (int order, int lens, int64_t nvals, CTF::Pair< dtype > pairs, bool with_vals, int64_t &str_len)
	serialize sparse tensor data to create string More...

template<typename dtype >
int64_t	read_data_mpiio (CTF::World const dw, char const fpath, char ***datastr)
	read sparse tensor data from file using MPI-I/O, creating string with one entry per line (different entries on each process) More...

template<typename dtype >
void	write_data_mpiio (CTF::World const dw, char const fpath, char *datastr, int64_t str_len)
	write sparse tensor data to file using MPI-I/O, from string with one entry per line (different entries on each process) More...

template<typename dtype >
dtype	default_add (dtype a, dtype b)

template<typename dtype , void()(int, dtype const , dtype *) fxpy>
void	default_mxpy (void X, void Y, int n, MPI_Datatype d)

template<typename dtype >
void	default_fxpy (int n, dtype const X, dtype Y)

template<typename dtype >
MPI_Op	get_default_maddop ()

template<>
MPI_Op	get_default_maddop< char > ()

template<>
MPI_Op	get_default_maddop< bool > ()

template<>
MPI_Op	get_default_maddop< int > ()

template<>
MPI_Op	get_default_maddop< int64_t > ()

template<>
MPI_Op	get_default_maddop< unsigned int > ()

template<>
MPI_Op	get_default_maddop< uint64_t > ()

template<>
MPI_Op	get_default_maddop< float > ()

template<>
MPI_Op	get_default_maddop< double > ()

template<>
MPI_Op	get_default_maddop< long double > ()

template<>
MPI_Op	get_default_maddop< std::complex< float > > ()

template<>
MPI_Op	get_default_maddop< std::complex< double > > ()

template<typename dtype >
MPI_Op	get_maddop (void(fxpy)(int, dtype const , dtype *))

CTF_int::algstrct const *	get_float_ring ()

CTF_int::algstrct const *	get_double_ring ()

CTF_int::algstrct const *	get_int_ring ()

CTF_int::algstrct const *	get_int64_t_ring ()

template<typename dtype >
void	gemm_batch (char taA, char taB, int l, int m, int n, int k, dtype alpha, dtype const A, dtype const B, dtype beta, dtype *C)

template<typename dtype >
void	gemm (char tA, char tB, int m, int n, int k, dtype alpha, dtype const A, dtype const B, dtype beta, dtype *C)

template<>
void	default_axpy< float > (int n, float alpha, float const X, int incX, float Y, int incY)

template<>
void	default_axpy< double > (int n, double alpha, double const X, int incX, double Y, int incY)

template<>
void	default_axpy< std::complex< float > > (int n, std::complex< float > alpha, std::complex< float > const X, int incX, std::complex< float > Y, int incY)

template<>
void	default_axpy< std::complex< double > > (int n, std::complex< double > alpha, std::complex< double > const X, int incX, std::complex< double > Y, int incY)

template<>
void	default_scal< float > (int n, float alpha, float *X, int incX)

template<>
void	default_scal< double > (int n, double alpha, double *X, int incX)

template<>
void	default_scal< std::complex< float > > (int n, std::complex< float > alpha, std::complex< float > *X, int incX)

template<>
void	default_scal< std::complex< double > > (int n, std::complex< double > alpha, std::complex< double > *X, int incX)

template<>
void	default_coomm< float > (int m, int n, int k, float alpha, float const A, int const rows_A, int const cols_A, int nnz_A, float const B, float beta, float *C)

template<>
void	default_coomm< double > (int m, int n, int k, double alpha, double const A, int const rows_A, int const cols_A, int nnz_A, double const B, double beta, double *C)

template<>
void	default_coomm< std::complex< float > > (int m, int n, int k, std::complex< float > alpha, std::complex< float > const A, int const rows_A, int const cols_A, int nnz_A, std::complex< float > const B, std::complex< float > beta, std::complex< float > *C)

template<>
void	default_coomm< std::complex< double > > (int m, int n, int k, std::complex< double > alpha, std::complex< double > const A, int const rows_A, int const cols_A, int nnz_A, std::complex< double > const B, std::complex< double > beta, std::complex< double > *C)

template<typename dtype >
void	muladd_csrmm (int m, int n, int k, dtype alpha, dtype const A, int const JA, int const IA, int nnz_A, dtype const B, dtype beta, dtype *C)

template<typename dtype >
void	muladd_csrmultd (int m, int n, int k, dtype const A, int const JA, int const IA, int nnz_A, dtype const B, int const JB, int const IB, int nnz_B, dtype *C)

template<typename dtype >
dtype	default_mul (dtype a, dtype b)

template<typename dtype >
void	default_axpy (int n, dtype alpha, dtype const X, int incX, dtype Y, int incY)

template<typename dtype >
void	default_scal (int n, dtype alpha, dtype *X, int incX)

template<typename dtype >
void	default_gemm (char tA, char tB, int m, int n, int k, dtype alpha, dtype const A, dtype const B, dtype beta, dtype *C)

template<typename dtype >
dtype **	get_grp_ptrs (int64_t grp_sz, int64_t ngrp, dtype const *data)

template<>
void	default_gemm< float > (char tA, char tB, int m, int n, int k, float alpha, float const A, float const B, float beta, float *C)

template<>
void	default_gemm< double > (char tA, char tB, int m, int n, int k, double alpha, double const A, double const B, double beta, double *C)

template<>
void	default_gemm< std::complex< float > > (char tA, char tB, int m, int n, int k, std::complex< float > alpha, std::complex< float > const A, std::complex< float > const B, std::complex< float > beta, std::complex< float > *C)

template<>
void	default_gemm< std::complex< double > > (char tA, char tB, int m, int n, int k, std::complex< double > alpha, std::complex< double > const A, std::complex< double > const B, std::complex< double > beta, std::complex< double > *C)

template<typename dtype >
void	default_gemm_batch (char taA, char taB, int l, int m, int n, int k, dtype alpha, dtype const A, dtype const B, dtype beta, dtype *C)

template<>
void	default_gemm_batch< float > (char taA, char taB, int l, int m, int n, int k, float alpha, float const A, float const B, float beta, float *C)

template<>
void	default_gemm_batch< double > (char taA, char taB, int l, int m, int n, int k, double alpha, double const A, double const B, double beta, double *C)

template<>
void	default_gemm_batch< std::complex< float > > (char taA, char taB, int l, int m, int n, int k, std::complex< float > alpha, std::complex< float > const A, std::complex< float > const B, std::complex< float > beta, std::complex< float > *C)

template<>
void	default_gemm_batch< std::complex< double > > (char taA, char taB, int l, int m, int n, int k, std::complex< double > alpha, std::complex< double > const A, std::complex< double > const B, std::complex< double > beta, std::complex< double > *C)

template<typename dtype >
void	default_coomm (int m, int n, int k, dtype alpha, dtype const A, int const rows_A, int const cols_A, int nnz_A, dtype const B, dtype beta, dtype *C)

bool	try_mkl_coo_to_csr (int64_t nz, int nrow, char csr_vs, int csr_ja, int csr_ia, char const coo_vs, int const coo_rs, int const coo_cs, int el_size)

bool	try_mkl_csr_to_coo (int64_t nz, int nrow, char const csr_vs, int const csr_ja, int const csr_ia, char coo_vs, int coo_rs, int coo_cs, int el_size)

template<typename dtype >
void	seq_coo_to_csr (int64_t nz, int nrow, dtype csr_vs, int csr_ja, int csr_ia, dtype const coo_vs, int const coo_rs, int const coo_cs)

template<typename dtype >
void	seq_csr_to_coo (int64_t nz, int nrow, dtype const csr_vs, int const csr_ja, int const csr_ia, dtype coo_vs, int coo_rs, int coo_cs)

template<typename dtype >
void	def_coo_to_csr (int64_t nz, int nrow, dtype csr_vs, int csr_ja, int csr_ia, dtype const coo_vs, int const coo_rs, int const coo_cs)

template<typename dtype >
void	def_csr_to_coo (int64_t nz, int nrow, dtype const csr_vs, int const csr_ja, int const csr_ia, dtype coo_vs, int coo_rs, int coo_cs)

template<typename dtype >
dtype	default_addinv (dtype a)

template<typename dtype , bool is_ord>
std::enable_if< is_ord, dtype >::type	default_abs (dtype a)

template<typename dtype , bool is_ord>
std::enable_if<!is_ord, dtype >::type	default_abs (dtype a)

template<typename dtype , dtype(*)(dtype) abs>
void	char_abs (char const a, char b)

template<typename dtype , bool is_ord>
std::enable_if< is_ord, dtype >::type	default_min (dtype a, dtype b)

template<typename dtype , bool is_ord>
std::enable_if<!is_ord, dtype >::type	default_min (dtype a, dtype b)

template<typename dtype , bool is_ord>
std::enable_if< is_ord, dtype >::type	default_max_lim ()

template<typename dtype , bool is_ord>
std::enable_if<!is_ord, dtype >::type	default_max_lim ()

template<typename dtype , bool is_ord>
std::enable_if< is_ord, dtype >::type	default_min_lim ()

template<typename dtype , bool is_ord>
std::enable_if<!is_ord, dtype >::type	default_min_lim ()

template<typename dtype , bool is_ord>
std::enable_if< is_ord, dtype >::type	default_max (dtype a, dtype b)

template<typename dtype , bool is_ord>
std::enable_if<!is_ord, dtype >::type	default_max (dtype a, dtype b)

template<typename dtype >
MPI_Datatype	get_default_mdtype (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< bool > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< std::complex< double > > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< std::complex< long double > > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< char > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< int > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< int64_t > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< unsigned int > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< uint64_t > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< float > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< double > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< long double > (bool &is_custom)

template<>
MPI_Datatype	get_default_mdtype< std::complex< float > > (bool &is_custom)

template<typename dtype >
constexpr bool	get_default_is_ord ()

Idx_Tensor *	get_full_intm (Idx_Tensor &A, Idx_Tensor &B, std::vector< char > out_inds, bool create_dummy=false)

std::vector< char >	det_uniq_inds (std::vector< Term * > const operands, std::vector< char > const out_inds)

std::vector< Term * >	contract_down_terms (algstrct sr, char tscale, std::vector< Term * > operands, std::vector< char > out_inds, int terms_to_leave, bool est_time=false, double *cost=NULL)

void	operator-= (double &d, CTF_int::Term const &tsr)

void	operator+= (double &d, CTF_int::Term const &tsr)

void	operator-= (int64_t &d, CTF_int::Term const &tsr)

void	operator+= (int64_t &d, CTF_int::Term const &tsr)

CTF_int::Contract_Term	operator* (double const &d, CTF_int::Term const &tsr)

CTF_int::Contract_Term	operator* (int64_t const &i, CTF_int::Term const &tsr)

void	calc_dim (int order, int64_t size, int const edge_len, mapping const edge_map, int64_t vrt_sz, int vrt_edge_len, int *blk_edge_len)
	calculate the block-sizes of a tensor More...

int	get_distribution_size (int order)

int	comp_dim_map (mapping const map_A, mapping const map_B)
	compares two mappings More...

void	copy_mapping (int order, mapping const mapping_A, mapping mapping_B)
	copies mapping A to B More...

int	copy_mapping (int order_A, int order_B, int const idx_A, mapping const mapping_A, int const idx_B, mapping mapping_B, int make_virt=1)
	copies mapping A to B More...

int	map_tensor (int num_phys_dims, int tsr_order, int const tsr_edge_len, int const tsr_sym_table, int restricted, CommData phys_comm, int const comm_idx, int fill, mapping tsr_edge_map)
	map a tensor More...

int	check_self_mapping (tensor const tsr, int const idx_map)
	checks mapping in preparation for tensors scale, summ or contract More...

int	map_self_indices (tensor const tsr, int const idx_map)
	create virtual mapping for idx_maps that have repeating indices More...

int	map_symtsr (int tsr_order, int const tsr_sym_table, mapping tsr_edge_map)
	adjust a mapping to maintan symmetry More...

int	stretch_virt (int order, int stretch_factor, mapping *maps)
	stretch virtualization by a factor More...

topology *	get_phys_topo (CommData glb_comm, TOPOLOGY mach)
	get dimension and torus lengths of specified topology More...

std::vector< topology * >	get_all_topos (CommData cdt, int n_uf, int const uniq_fact, int const mults, int n_prepend, int const *prelens)
	computes all unique factorizations into non-primes each yielding a topology, prepending additional factors as specified More...

std::vector< topology * >	get_generic_topovec (CommData cdt)
	computes all topology configurations given undelying physical topology information More...

std::vector< topology * >	peel_perm_torus (topology *phys_topology, CommData cdt)
	folds specified topology and all of its permutations into all configurations of lesser dimensionality More...

std::vector< topology * >	peel_torus (topology const *topo, CommData glb_comm)
	folds specified topology into all configurations of lesser dimensionality More...

int	find_topology (topology const topo, std::vector< topology > &topovec)
	searches for an equivalent topology in avector of topologies More...

int	get_best_topo (int64_t nvirt, int topo, CommData global_comm, int64_t bcomm_vol=0, int64_t bmemuse=0)
	get the best topologoes (least nvirt) over all procs More...

void	extract_free_comms (topology const topo, int order_A, mapping const edge_map_A, int order_B, mapping const edge_map_B, int &num_sub_phys_dims, CommData psub_phys_comm, int *pcomm_idx)
	extracts the set of physical dimensions still available for mapping More...

int	can_morph (topology const topo_keep, topology const topo_change)
	determines if two topologies are compatible with each other More...

void	morph_topo (topology const new_topo, topology const old_topo, int order, mapping *edge_map)
	morphs a tensor topology into another More...

void	pad_cyclic_pup_virt_buff (int const sym, distribution const &old_dist, distribution const &new_dist, int const len, int const old_phys_dim, int const old_phys_edge_len, int const old_virt_edge_len, int64_t old_virt_nelem, int const old_offsets, int const old_permutation, int total_np, int const new_phys_dim, int const new_phys_edge_len, int const new_virt_edge_len, int64_t new_virt_nelem, char old_data, char *new_data, int forward, int const bucket_offset, char const alpha, char const beta, algstrct const sr)

void	cyclic_reshuffle (int const sym, distribution const &old_dist, int const old_offsets, int const old_permutation, distribution const &new_dist, int const new_offsets, int const new_permutation, char tsr_data, char tsr_cyclic_data, algstrct const sr, CommData ord_glb_comm, bool reuse_buffers, char const alpha, char const beta)
	Goes from any set of phases to any new set of phases. More...

template<int idim>
void	redist_bucket (int const bucket_offset, int64_t const data_offset, int const ivmax_pre, int rep_phase0, int virt_dim0, bool data_to_buckets, char __restrict__ data, char __restrict__ buckets, int64_t counts, algstrct const *sr, int64_t data_off, int bucket_off, int prev_idx)

template<>
void	redist_bucket< 0 > (int const bucket_offset, int64_t const data_offset, int const ivmax_pre, int rep_phase0, int virt_dim0, bool data_to_buckets, char __restrict__ data, char __restrict__ buckets, int64_t counts, algstrct const *sr, int64_t data_off, int bucket_off, int prev_idx)

void	redist_bucket_r0 (int const bucket_offset, int64_t const data_offset, int const ivmax_pre, int rep_phase0, int rep_idx0, int virt_dim0, bool data_to_buckets, char __restrict__ data, char __restrict__ buckets, int64_t counts, algstrct const *sr, int64_t data_off, int bucket_off, int prev_idx)

int	get_glb (int i, int s, int t)

int	get_loc (int g, int s, int t)

template<int idim>
int64_t	calc_cnt (int const sym, int const rep_phase, int const sphase, int const gidx_off, int const edge_len, int const loc_edge_len)
	computes the cardinality of the set of elements of a tensor of order idim+1 that are owned by processor index gidx_off in a distribution with dimensions sphase More...

template<>
int64_t	calc_cnt< 0 > (int const sym, int const rep_phase, int const sphase, int const gidx_off, int const edge_len, int const loc_edge_len)

template<int idim>
int64_t *	calc_sy_pfx (int const sym, int const rep_phase, int const sphase, int const gidx_off, int const edge_len, int const loc_edge_len)
	computes the cardinality of the sets of elements of a tensor of order idim+1 for different values of the idim'th tensor dimension More...

template<>
int64_t *	calc_sy_pfx< 1 > (int const sym, int const rep_phase, int const sphase, int const gidx_off, int const edge_len, int const loc_edge_len)

template<int idim>
void	calc_drv_cnts (int order, int const sym, int64_t counts, int const rep_phase, int const rep_phase_lda, int const sphase, int const phys_phase, int gidx_off, int const edge_len, int const *loc_edge_len)

template<>
void	calc_drv_cnts< 0 > (int order, int const sym, int64_t counts, int const rep_phase, int const rep_phase_lda, int const sphase, int const phys_phase, int gidx_off, int const edge_len, int const *loc_edge_len)

template<int idim>
void	calc_cnt_from_rep_cnt (int const rep_phase, int const pe_offset, int const bucket_offset, int64_t const old_counts, int64_t *counts, int bucket_off, int pe_off, int dir)

template<>
void	calc_cnt_from_rep_cnt< 0 > (int const rep_phase, int const pe_offset, int const bucket_offset, int64_t const old_counts, int64_t *counts, int bucket_off, int pe_off, int dir)

void	calc_drv_displs (int const sym, int const edge_len, distribution const &old_dist, distribution const &new_dist, int64_t *counts, int idx_lyr)

void	precompute_offsets (distribution const &old_dist, distribution const &new_dist, int const sym, int const len, int const rep_phase, int const phys_edge_len, int const virt_edge_len, int const virt_dim, int const virt_lda, int64_t virt_nelem, int pe_offset, int bucket_offset, int64_t data_offset, int *ivmax_pre)

double	dgtog_est_time (int64_t tot_sz, int np)
	estimates execution time, given this processor sends a receives tot_sz across np procs More...

void	dgtog_reshuffle (int const sym, int const edge_len, distribution const &old_dist, distribution const &new_dist, char ptr_tsr_data, char ptr_tsr_new_data, algstrct const *sr, CommData ord_glb_comm)

void	glb_ord_pup (int const sym, distribution const &old_dist, distribution const &new_dist, int const len, int const old_phys_dim, int const old_phys_edge_len, int const old_virt_edge_len, int64_t old_virt_nelem, int const old_offsets, int const old_permutation, int total_np, int const new_phys_dim, int const new_phys_edge_len, int const new_virt_edge_len, int64_t new_virt_nelem, char old_data, char *new_data, int forward, int const bucket_offset, char const alpha, char const beta, algstrct const sr)

template<int idim>
void	ord_glb (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int prev_idx=0, int64_t glb_ord_offset=0, int64_t blk_ord_offset=0)

template<>
void	ord_glb< 0 > (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int prev_idx, int64_t glb_ord_offset, int64_t blk_ord_offset)

template void	ord_glb< 7 > (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int prev_idx, int64_t glb_ord_offset, int64_t blk_ord_offset)

template<int idim>
void	ord_glb_omp (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int const idx_st, int const idx_end, int prev_idx=0, int64_t glb_ord_offset=0, int64_t blk_ord_offset=0)

template<>
void	ord_glb_omp< 0 > (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int const idx_st, int const idx_end, int prev_idx, int64_t glb_ord_offset, int64_t blk_ord_offset)

template void	ord_glb_omp< 7 > (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr, int const idx_st, int const idx_end, int prev_idx, int64_t glb_ord_offset, int64_t blk_ord_offset)

void	order_globally (int const sym, distribution const &dist, int const virt_edge_len, int const virt_phase_lda, int64_t vbs, bool dir, char const tsr_data_in, char tsr_data_out, algstrct const sr)
	reorder local buffer so that elements are in ordered according to where they are in the global tensor (interleave virtual blocks) More...

char *	glb_cyclic_reshuffle (int const sym, distribution const &old_dist, int const old_offsets, int const old_permutation, distribution const &new_dist, int const new_offsets, int const new_permutation, char ptr_tsr_data, char ptr_tsr_cyclic_data, algstrct const sr, CommData ord_glb_comm, bool reuse_buffers, char const alpha, char const beta)
	Goes from any set of phases to any new set of phases. More...

bool	hptt_is_applicable (int order, int const *new_order, int elementSize)
	Checks if the HPTT library is applicable. More...

void	nosym_transpose_hptt (int order, int const st_new_order, int const st_edge_len, int dir, char const st_buffer, char new_buffer, algstrct const *sr)

void	nosym_transpose (tensor A, int all_fdim_A, int const all_flen_A, int const *new_order, int dir)

void	nosym_transpose (int order, int const new_order, int const edge_len, char data, int dir, algstrct const sr)
	transposes a non-symmetric (folded) tensor More...

void	nosym_transpose (int order, int const new_order, int const edge_len, char const data, int dir, int max_ntd, char tswap_data, int64_t chunk_size, algstrct const *sr)
	transposes a non-symmetric (folded) tensor internal kernel More...

double	est_time_transp (int order, int const new_order, int const edge_len, int dir, algstrct const *sr)
	estimates time needed to transposes a non-symmetric (folded) tensor based on performance models More...

void	nosym_transpose_hptt (int order, int const edge_len, int dir, tensor &A)
	High-performance implementation of nosym_transpose using HPTT. More...

void	pad_key (int order, int64_t num_pair, int const edge_len, int const padding, PairIterator pairs, algstrct const sr, int const offsets=NULL)
	applies padding to keys More...

void	depad_tsr (int order, int64_t num_pair, int const edge_len, int const sym, int const padding, int const prepadding, char const pairsb, char new_pairsb, int64_t new_num_pair, algstrct const sr)
	retrieves the unpadded pairs More...

void	zero_padding (int order, int64_t size, int nvirt, int const edge_len, int const sym, int const padding, int const phase, int const phys_phase, int const virt_phase, int const cphase_rank, char vdata, algstrct const *sr)
	sets to zero all values in padded region of tensor More...

void	scal_diag (int order, int64_t size, int nvirt, int const edge_len, int const sym, int const padding, int const phase, int const phys_phase, int const virt_phase, int const cphase_rank, char vdata, algstrct const sr, int const sym_mask)
	scales each element by 1/(number of entries equivalent to it after permutation of indices for which sym_mask is 1) More...

void	padded_reshuffle (int const sym, distribution const &old_dist, distribution const &new_dist, char tsr_data, char *tsr_cyclic_data, algstrct const sr, CommData ord_glb_comm)
	Reshuffle elements using key-value pair read/write. More...

int **	compute_bucket_offsets (distribution const &old_dist, distribution const &new_dist, int const len, int const old_phys_edge_len, int const old_virt_lda, int const old_offsets, int const old_permutation, int const new_phys_edge_len, int const new_virt_lda, int forward, int old_virt_np, int new_virt_np, int const *old_virt_edge_len)
	computes offsets for redistribution targets along each edge length More...

void	calc_cnt_displs (int const sym, distribution const &old_dist, distribution const &new_dist, int new_nvirt, int np, int const old_virt_edge_len, int const new_virt_lda, int64_t send_counts, int64_t recv_counts, int64_t send_displs, int64_t recv_displs, CommData ord_glb_comm, int idx_lyr, int const *bucket_offset)
	assigns keys to an array of values More...

double	blres_est_time (int64_t tot_sz, int nv0, int nv1)
	estimates execution time, given this processor sends a receives tot_sz across np procs More...

void	block_reshuffle (distribution const &old_dist, distribution const &new_dist, char tsr_data, char &tsr_cyclic_data, algstrct const *sr, CommData glb_comm)
	Reshuffle elements by block given the global phases stay the same. More...

int	can_block_reshuffle (int order, int const old_phase, mapping const map)
	determines if tensor can be permuted by block More...

void	permute_keys (int order, int num_pair, int const edge_len, int const new_edge_len, int const permutation, char pairs, int64_t new_num_pair, algstrct const *sr)
	permutes keys More...

void	depermute_keys (int order, int num_pair, int const edge_len, int const new_edge_len, int const permutation, char pairs, algstrct const sr)
	depermutes keys (apply P^T) More...

void	assign_keys (int order, int64_t size, int nvirt, int const edge_len, int const sym, int const phase, int const phys_phase, int const virt_dim, int phase_rank, char const vdata, char vpairs, algstrct const *sr)
	assigns keys to an array of values More...

void	spsfy_tsr (int order, int64_t size, int nvirt, int const edge_len, int const sym, int const phase, int const phys_phase, int const virt_dim, int phase_rank, char const vdata, char &vpairs, int64_t nnz_blk, algstrct const sr, int64_t const edge_lda, std::function< bool(char const )> f)
	extracts all tensor values (in pair format) that pass a sparsifier function (including padded zeros if they pass the fliter) More...

void	bucket_by_pe (int order, int64_t num_pair, int64_t np, int const phys_phase, int const virt_phase, int const bucket_lda, int const edge_len, ConstPairIterator mapped_data, int64_t bucket_counts, int64_t bucket_off, PairIterator bucket_data, algstrct const *sr)
	buckets key-value pairs by processor according to distribution More...

int64_t *	bucket_by_virt (int order, int num_virt, int64_t num_pair, int const phys_phase, int const virt_phase, int const edge_len, ConstPairIterator mapped_data, PairIterator bucket_data, algstrct const sr)
	buckets key value pairs by block/virtual-processor More...

void	readwrite (int order, int64_t size, char const alpha, char const beta, int nvirt, int const edge_len, int const sym, int const phase, int const phys_phase, int const virt_dim, int phase_rank, char vdata, char pairs, char rw, algstrct const *sr)
	read or write pairs from / to tensor More...

void	wr_pairs_layout (int order, int np, int64_t inwrite, char const alpha, char const beta, char rw, int num_virt, int const sym, int const edge_len, int const padding, int const phase, int const phys_phase, int const virt_phase, int virt_phys_rank, int const bucket_lda, char wr_pairs_buf, char rw_data, CommData glb_comm, algstrct const sr, bool is_sparse, int64_t nnz_loc, int64_t nnz_blk, char *&pprs_new, int64_t &nnz_loc_new)
	read or write pairs from / to tensor More...

void	read_loc_pairs (int order, int64_t nval, int num_virt, int const sym, int const edge_len, int const padding, int const phase, int const phys_phase, int const virt_phase, int phase_rank, int64_t nread, char const data, char pairs, algstrct const sr)
	read tensor pairs local to processor More...

void	sp_read (algstrct const sr, int64_t ntsr, ConstPairIterator prs_tsr, char const alpha, int64_t nread, PairIterator prs_read, char const *beta)
	reads elements of a sparse set defining the tensor, into a sparse read set with potentially repeating keys More...

void	sp_write (int num_virt, algstrct const sr, int64_t vntsr, ConstPairIterator vprs_tsr, char const beta, int64_t vnwrite, ConstPairIterator vprs_write, char const alpha, int64_t vnnew, char *&pprs_new)
	writes pairs in a sparse write set to the sparse set of elements defining the tensor, resulting in a set of size between ntsr and ntsr+nwrite More...

void	inv_idx (int order_A, int const idx_A, int order_tot, int **idx_arr)
	invert index map More...

int	strip_diag (int order, int order_tot, int const idx_map, int64_t vrt_sz, mapping const edge_map, topology const topo, algstrct const sr, int blk_edge_len, int64_t blk_sz, strp_tsr **stpr)
	build stack required for stripping out diagonals of tensor More...

int	sym_seq_scl_ref (char const alpha, char A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A)
	performs symmetric scaling using algstrct const * sr_A More...

int	sym_seq_scl_cust (char const alpha, char A, algstrct const sr_A, int const order_A, int const edge_len_A, int const sym_A, int const idx_map_A, endomorphism const *func)
	performs symmetric scaling using custom func More...

void	inc_tot_mem_used (int64_t a)

void	set_mem_size (int64_t size)
	sets what fraction of the memory capacity CTF can use More...

void	set_memcap (double cap)
	sets what fraction of the memory capacity CTF can use More...

std::list< mem_transfer >	contract_mst ()
	gets rid of empty space on the stack More...

std::list< mem_loc > *	get_mst ()

void	mst_create (int64_t size)
	initializes stack buffer More...

void	mem_create ()
	create instance of memory manager More...

void	mem_exit (int rank)
	exit instance of memory manager More...

int	mst_free (void *ptr)
	frees buffer allocated on stack More...

int	untag_mem (void *ptr)
	stops tracking memory allocated by CTF, so user doesn't have to call free More...

int	cdealloc (void *ptr, int const tid)
	free abstraction More...

int	cdealloc_cond (void *ptr)
	free abstraction (conditional (no error if not found)) More...

int	get_num_instances ()

int64_t	proc_bytes_used ()
	gives total memory used on this MPI process More...

int64_t	proc_bytes_total ()
	gives total memory size per MPI process More...

int64_t	proc_bytes_available ()
	gives total memory available on this MPI process More...

std::vector< Model * > &	get_all_models ()

void	print_all_models ()

void	load_all_models (std::string file_name)

void	write_all_models (std::string file_name)

void	dump_all_models (std::string path)

double	cddot (int n, const double dX, int incX, const double dY, int incY)

void	cdgeqrf (int const M, int const N, double A, int const LDA, double TAU2, double WORK, int const LWORK, int INFO)

void	cdormqr (char SIDE, char TRANS, int M, int N, int K, double const A, int LDA, double const TAU2, double C, int LDC, double WORK, int LWORK, int *INFO)

void	cdgelsd (int m, int n, int k, double const A, int lda_A, double B, int lda_B, double S, double cond, int rank, double work, int lwork, int iwork, int *info)

template<int nparam>
bool	comp_time_param (const time_param< nparam > &a, const time_param< nparam > &b)

void	offload_init ()
	initialize offloading, e.g. create cublas More...

void	offload_exit ()
	exit offloading, e.g. destroy cublas More...

double	estimate_download_time (int64_t size)
	estimate time it takes to upload More...

double	estimate_upload_time (int64_t size)
	estimate time it takes to download More...

void	host_pinned_alloc (void **ptr, int64_t size)
	allocate a pinned host buffer More...

void	host_pinned_free (void *ptr)
	free a pinned host buffer More...

template<typename dtype >
void	offload_gemm (char tA, char tB, int m, int n, int k, dtype alpha, offload_tsr &A, int lda_A, offload_tsr &B, int lda_B, dtype beta, offload_tsr &C, int lda_C)

template<typename dtype >
void	offload_gemm (char tA, char tB, int m, int n, int k, dtype alpha, dtype const dev_A, int lda_A, dtype const dev_B, int lda_B, dtype beta, dtype *dev_C, int lda_C)

void	calc_idx_arr (int order, int const lens, int const sym, int64_t idx, int *idx_arr)

void	sy_calc_idx_arr (int order, int const lens, int const sym, int64_t idx, int *idx_arr)
	same as above except assumes sym only NS or SY More...

void	permute (int order, int const perm, int arr)
	permute an array More...

void	permute_target (int order, int const perm, int arr)
	permutes a permutation array More...

void	socopy (int64_t m, int64_t n, int64_t lda_a, int64_t lda_b, int64_t const sizes_a, int64_t &sizes_b, int64_t *&offsets_b)

void	spcopy (int64_t m, int64_t n, int64_t lda_a, int64_t lda_b, int64_t const sizes_a, int64_t const offsets_a, char const a, int64_t const sizes_b, int64_t const offsets_b, char b)

int64_t	fact (int64_t n)

int64_t	choose (int64_t n, int64_t k)

void	get_choice (int64_t n, int64_t k, int64_t ch, int *chs)

int64_t	chchoose (int64_t n, int64_t k)

int64_t	getTotalSystemMemory ()

int	free_cond (void *ptr)

int	gcd (int a, int b)

int	lcm (int a, int b)

void	lda_cpy (int el_size, int nrow, int ncol, int lda_A, int lda_B, const char A, char B)
	Copies submatrix to submatrix (column-major) More...

void	coalesce_bwd (int el_size, char B, char const B_aux, int k, int n, int kb)
	we receive a contiguous buffer kb-by-n B and (k-kb)-by-n B_aux which is the block below. To get a k-by-n buffer, we need to combine this buffer with our original block. Since we are working with column-major ordering we need to interleave the blocks. Thats what this function does. More...

int64_t	get_coo_size (int64_t nnz, int val_size)

int64_t	get_csr_size (int64_t nnz, int nrow, int val_size)
	computes the size of a serialized CSR matrix More...

template<int idim>
void	spA_dnB_seq_sum_loop (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, char const beta, char &B, algstrct const sr_B, int order_B, int64_t idx_B, int const edge_len_B, int64_t const lda_B, int const sym_B, univar_function const *func)

template<>
void	spA_dnB_seq_sum_loop< 0 > (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, char const beta, char &B, algstrct const sr_B, int order_B, int64_t idx_B, int const edge_len_B, int64_t const lda_B, int const sym_B, univar_function const *func)

template void	spA_dnB_seq_sum_loop< MAX_ORD > (char const alpha, ConstPairIterator &A, int64_t &size_A, algstrct const sr_A, char const beta, char &B, algstrct const sr_B, int order_B, int64_t idx_B, int const edge_len_B, int64_t const lda_B, int const sym_B, univar_function const *func)

void	spA_dnB_seq_sum (char const alpha, char const A, int64_t size_A, algstrct const sr_A, char const beta, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, univar_function const *func)
	performs summation between two sparse tensors assumes A contains key value pairs sorted by key, with index permutation preapplied and with no repeated indices More...

void	dnA_spB_seq_sum (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, char const beta, char const B, int64_t size_B, char &new_B, int64_t &new_size_B, algstrct const sr_B, univar_function const func)
	performs summation between two sparse tensors assumes B contain key value pairs sorted by key, with index permutation preapplied and with no repeated indices More...

void	spspsum (algstrct const sr_A, int64_t nA, ConstPairIterator prs_A, char const beta, algstrct const sr_B, int64_t nB, ConstPairIterator prs_B, char const alpha, int64_t &nnew, char &pprs_new, univar_function const func, int64_t map_pfx)
	As pairs in a sparse A set to the sparse set of elements defining the tensor, resulting in a set of size between nB and nB+nA. More...

void	spA_spB_seq_sum (char const alpha, char const A, int64_t size_A, algstrct const sr_A, char const beta, char B, int64_t size_B, char &new_B, int64_t &new_size_B, algstrct const sr_B, univar_function const func, int64_t map_pfx)
	performs summation between two sparse tensors assumes A and B contain key value pairs sorted by key, with index permutation preapplied and with no repeated indices More...

void	inv_idx (int order_A, int const idx_A, int order_B, int const idx_B, int order_tot, int *idx_arr)
	invert index map More...

template<int idim>
void	sym_seq_sum_loop (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, univar_function const func, int const idx, int const rev_idx_map, int idx_max)

template<>
void	sym_seq_sum_loop< 0 > (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, univar_function const func, int const idx, int const rev_idx_map, int idx_max)

template void	sym_seq_sum_loop< MAX_ORD > (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, uint64_t const offsets_A, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, uint64_t const offsets_B, univar_function const func, int const idx, int const rev_idx_map, int idx_max)

void	compute_syoffs (algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, int tot_order, int const rev_idx_map, uint64_t &offsets_A, uint64_t *&offsets_B)

int	sym_seq_sum_ref (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const beta, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B)
	performs symmetric contraction with unblocked reference kernel More...

int	sym_seq_sum_inr (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const beta, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, int inr_stride)
	performs symmetric summation with blocked daxpy More...

int	sym_seq_sum_cust (char const alpha, char const A, algstrct const sr_A, int order_A, int const edge_len_A, int const sym_A, int const idx_map_A, char const beta, char B, algstrct const sr_B, int order_B, int const edge_len_B, int const sym_B, int const idx_map_B, univar_function const *func)
	performs symmetric summation with custom elementwise function More...

void	desymmetrize (tensor sym_tsr, tensor nonsym_tsr, bool is_C)
	unfolds the data of a tensor More...

void	symmetrize (tensor sym_tsr, tensor nonsym_tsr)
	folds the data of a tensor More...

void	cmp_sym_perms (int ndim, int const sym, int nperm, int *perm, double sign)
	finds all permutations of a tensor according to a symmetry More...

void	order_perm (tensor const A, tensor const B, int idx_arr, int off_A, int off_B, int idx_A, int *idx_B, int &add_sign, int &mod)
	orders the summation indices of one tensor that don't break summation symmetries More...

void	order_perm (tensor const A, tensor const B, tensor const C, int idx_arr, int off_A, int off_B, int off_C, int idx_A, int idx_B, int *idx_C, int &add_sign, int &mod)
	orders the contraction indices of one tensor that don't break contraction symmetries More...

void	add_sym_perm (std::vector< summation > &perms, std::vector< int > &signs, summation const &new_perm, int new_sign)
	puts a summation map into a nice ordering according to preserved symmetries, and adds it if it is distinct More...

void	add_sym_perm (std::vector< contraction > &perms, std::vector< int > &signs, contraction const &new_perm, int new_sign)
	puts a contraction map into a nice ordering according to preserved symmetries, and adds it if it is distinct More...

void	get_sym_perms (summation const &sum, std::vector< summation > &perms, std::vector< int > &signs)
	finds all permutations of a summation that must be done for a broken symmetry More...

void	get_sym_perms (contraction const &ctr, std::vector< contraction > &perms, std::vector< int > &signs)
	finds all permutations of a contraction that must be done for a broken symmetry More...

void	depin (algstrct const sr, int order, int const lens, int const divisor, int nvirt, int const virt_dim, int const phys_rank, char X, int64_t &new_nnz_B, int64_t nnz_blk, char &new_B, bool check_padding)
	depins keys of n pairs More...

double	spredist_est_time (int64_t size, int np)

template<typename dtype >
void	abs_helper (tensor A, tensor B)

template<typename dtype >
void	pow_helper (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template<typename dtype >
void	all_helper (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template<typename dtype >
void	conj_helper (tensor A, tensor B)

template<typename dtype >
void	get_real (tensor A, tensor B)

template<typename dtype >
void	get_imag (tensor A, tensor B)

template<typename dtype >
void	set_real (tensor A, tensor B)

template<typename dtype >
void	set_imag (tensor A, tensor B)

template<typename dtype >
void	any_helper (tensor A, tensor B_bool, char const idx_A, char const idx_B)

int64_t	sum_bool_tsr (tensor *A)
	sum all 1 values in boolean tensor More...

void	subsample (tensor *A, double probability)
	extract a sample of the entries (if sparse of the current nonzeros) More...

void	matrix_qr (tensor A, tensor Q, tensor *R)

void	matrix_qr_cmplx (tensor A, tensor Q, tensor *R)

void	matrix_svd (tensor A, tensor U, tensor S, tensor VT, int rank)

void	matrix_svd_cmplx (tensor A, tensor U, tensor S, tensor VT, int rank)

void	conv_type (int type_idx1, int type_idx2, tensor A, tensor B)
	convert tensor from one type to another More...

template void	conj_helper< float > (tensor A, tensor B)

template void	conj_helper< double > (tensor A, tensor B)

template void	set_real< float > (tensor A, tensor B)

template void	set_imag< float > (tensor A, tensor B)

template void	set_real< double > (tensor A, tensor B)

template void	set_imag< double > (tensor A, tensor B)

template void	get_real< float > (tensor A, tensor B)

template void	get_imag< float > (tensor A, tensor B)

template void	get_real< double > (tensor A, tensor B)

template void	get_imag< double > (tensor A, tensor B)

template void	tensor::compare_elementwise< std::complex< double > > (tensor A, tensor B)

template void	tensor::compare_elementwise< std::complex< float > > (tensor A, tensor B)

template void	abs_helper< std::complex< double > > (tensor A, tensor B)

template void	abs_helper< std::complex< float > > (tensor A, tensor B)

template void	abs_helper< double > (tensor A, tensor B)

template void	abs_helper< float > (tensor A, tensor B)

template void	abs_helper< int64_t > (tensor A, tensor B)

template void	abs_helper< bool > (tensor A, tensor B)

template void	abs_helper< int32_t > (tensor A, tensor B)

template void	abs_helper< int16_t > (tensor A, tensor B)

template void	abs_helper< int8_t > (tensor A, tensor B)

template void	pow_helper< std::complex< double > > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< std::complex< float > > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< double > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< float > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< int64_t > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< bool > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< int32_t > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< int16_t > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	pow_helper< int8_t > (tensor A, tensor B, tensor C, char const idx_A, char const idx_B, char const idx_C)

template void	all_helper< std::complex< double > > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< std::complex< float > > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< int64_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< double > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< float > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< bool > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< int32_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< int16_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	all_helper< int8_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< std::complex< double > > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< std::complex< float > > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< double > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< float > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< int64_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< bool > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< int32_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< int16_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

template void	any_helper< int8_t > (tensor A, tensor B_bool, char const idx_A, char const idx_B)

Variables
LinModel< 3 >	seq_tsr_ctr_mdl_cst (seq_tsr_ctr_mdl_cst_init,"seq_tsr_ctr_mdl_cst")

LinModel< 3 >	seq_tsr_ctr_mdl_ref (seq_tsr_ctr_mdl_ref_init,"seq_tsr_ctr_mdl_ref")

LinModel< 3 >	seq_tsr_ctr_mdl_inr (seq_tsr_ctr_mdl_inr_init,"seq_tsr_ctr_mdl_inr")

LinModel< 3 >	seq_tsr_ctr_mdl_off (seq_tsr_ctr_mdl_off_init,"seq_tsr_ctr_mdl_off")

LinModel< 3 >	seq_tsr_ctr_mdl_cst_inr (seq_tsr_ctr_mdl_cst_inr_init,"seq_tsr_ctr_mdl_cst_inr")

LinModel< 3 >	seq_tsr_ctr_mdl_cst_off (seq_tsr_ctr_mdl_cst_off_init,"seq_tsr_ctr_mdl_cst_off")

LinModel< 3 >	seq_tsr_spctr_cst_off_k0 (seq_tsr_spctr_cst_off_k0_init,"seq_tsr_spctr_cst_off_k0")

LinModel< 3 >	seq_tsr_spctr_cst_off_k1 (seq_tsr_spctr_cst_off_k1_init,"seq_tsr_spctr_cst_off_k1")

LinModel< 3 >	seq_tsr_spctr_cst_off_k2 (seq_tsr_spctr_cst_off_k2_init,"seq_tsr_spctr_cst_off_k2")

LinModel< 3 >	seq_tsr_spctr_off_k0 (seq_tsr_spctr_off_k0_init,"seq_tsr_spctr_off_k0")

LinModel< 3 >	seq_tsr_spctr_off_k1 (seq_tsr_spctr_off_k1_init,"seq_tsr_spctr_off_k1")

LinModel< 3 >	seq_tsr_spctr_off_k2 (seq_tsr_spctr_off_k2_init,"seq_tsr_spctr_off_k2")

LinModel< 3 >	seq_tsr_spctr_cst_k0 (seq_tsr_spctr_cst_k0_init,"seq_tsr_spctr_cst_k0")

LinModel< 3 >	seq_tsr_spctr_cst_k1 (seq_tsr_spctr_cst_k1_init,"seq_tsr_spctr_cst_k1")

LinModel< 3 >	seq_tsr_spctr_cst_k2 (seq_tsr_spctr_cst_k2_init,"seq_tsr_spctr_cst_k2")

LinModel< 3 >	seq_tsr_spctr_cst_k3 (seq_tsr_spctr_cst_k3_init,"seq_tsr_spctr_cst_k3")

LinModel< 3 >	seq_tsr_spctr_cst_k4 (seq_tsr_spctr_cst_k4_init,"seq_tsr_spctr_cst_k4")

LinModel< 3 >	seq_tsr_spctr_k0 (seq_tsr_spctr_k0_init,"seq_tsr_spctr_k0")

LinModel< 3 >	seq_tsr_spctr_k1 (seq_tsr_spctr_k1_init,"seq_tsr_spctr_k1")

LinModel< 3 >	seq_tsr_spctr_k2 (seq_tsr_spctr_k2_init,"seq_tsr_spctr_k2")

LinModel< 3 >	seq_tsr_spctr_k3 (seq_tsr_spctr_k3_init,"seq_tsr_spctr_k3")

LinModel< 3 >	seq_tsr_spctr_k4 (seq_tsr_spctr_k4_init,"seq_tsr_spctr_k4")

LinModel< 2 >	pin_keys_mdl (pin_keys_mdl_init,"pin_keys_mdl")

std::mersenne_twister_engine< std::uint_fast64_t, 64, 312, 156, 31, 0xb5026f5aa96619e9, 29, 0x5555555555555555, 17, 0x71d67fffeda60000, 37, 0xfff7eee000000000, 43, 6364136223846793005 >	rng

LinModel< 3 >	alltoall_mdl (alltoall_mdl_init,"alltoall_mdl")

LinModel< 3 >	alltoallv_mdl (alltoallv_mdl_init,"alltoallv_mdl")

LinModel< 3 >	red_mdl (red_mdl_init,"red_mdl")

LinModel< 3 >	red_mdl_cst (red_mdl_cst_init,"red_mdl_cst")

LinModel< 3 >	allred_mdl (allred_mdl_init,"allred_mdl")

LinModel< 3 >	allred_mdl_cst (allred_mdl_cst_init,"allred_mdl_cst")

LinModel< 3 >	bcast_mdl (bcast_mdl_init,"bcast_mdl")

int64_t	total_flop_count = 0

MPI_Datatype	MPI_CTF_DOUBLE_COMPLEX = MPI_CXX_DOUBLE_COMPLEX

CTF::Ring< float >	float_ring = CTF::Ring<float>()

CTF::Ring< double >	double_ring = CTF::Ring<double>()

CTF::Ring< int >	int_ring = CTF::Ring<int>()

CTF::Ring< int64_t >	int64_t_ring = CTF::Ring<int64_t>()

MPI_Datatype	MPI_CTF_BOOL = MPI_CXX_BOOL

MPI_Datatype	MPI_CTF_LONG_DOUBLE_COMPLEX = MPI_CXX_LONG_DOUBLE_COMPLEX

std::set< grid_wrapper >	scalapack_grids
	index for ScaLAPACK processor grids More...

LinModel< 3 >	dgtog_res_mdl (dgtog_res_mdl_init,"dgtog_res_mdl")

LinModel< 2 >	long_contig_transp_mdl (long_contig_transp_mdl_init,"long_contig_transp_mdl")

LinModel< 2 >	shrt_contig_transp_mdl (shrt_contig_transp_mdl_init,"shrt_contig_transp_mdl")

LinModel< 2 >	non_contig_transp_mdl (non_contig_transp_mdl_init,"non_contig_transp_mdl")

LinModel< 2 >	blres_mdl (blres_mdl_init,"blres_mdl")

double	seq_tsr_spctr_cst_off_k0_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_cst_off_k1_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_cst_off_k2_init [] = {-2.1996E-04, 3.1883E-09, 3.8743E-11}

double	seq_tsr_spctr_off_k0_init [] = {8.6970E-06, 4.5598E-11, 1.1544E-09}

double	seq_tsr_spctr_off_k1_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_off_k2_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_cst_k0_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_cst_k1_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_cst_k2_init [] = {-8.8459E-08, 8.1207E-10, -2.8486E-12}

double	seq_tsr_spctr_cst_k3_init [] = {1.8504E-08, 2.9154E-11, 2.1973E-11}

double	seq_tsr_spctr_cst_k4_init [] = {2.0948E-05, 1.2294E-09, 8.0037E-10}

double	seq_tsr_spctr_k0_init [] = {2.2620E-08, -5.7494E-10, 2.2146E-09}

double	seq_tsr_spctr_k1_init [] = {5.3745E-06, 3.6464E-08, 2.2334E-10}

double	seq_tsr_spctr_k2_init [] = {3.0917E-08, 5.2181E-11, 4.1634E-12}

double	seq_tsr_spctr_k3_init [] = {7.2456E-08, 1.5128E-10, -1.5528E-12}

double	seq_tsr_spctr_k4_init [] = {1.6880E-07, 4.9411E-10, 9.2847E-13}

double	pin_keys_mdl_init [] = {3.1189E-09, 6.6717E-08}

double	seq_tsr_ctr_mdl_cst_init [] = {5.1626E-06, -6.3215E-11, 3.9638E-09}

double	seq_tsr_ctr_mdl_ref_init [] = {4.9138E-08, 5.8290E-10, 4.8575E-11}

double	seq_tsr_ctr_mdl_inr_init [] = {2.0647E-08, 1.9721E-10, 2.9948E-11}

double	seq_tsr_ctr_mdl_off_init [] = {6.2925E-05, 1.7449E-11, 1.7211E-12}

double	seq_tsr_ctr_mdl_cst_inr_init [] = {1.3863E-04, 2.0119E-10, 9.8820E-09}

double	seq_tsr_ctr_mdl_cst_off_init [] = {8.4844E-04, -5.9246E-11, 3.5247E-10}

double	long_contig_transp_mdl_init [] = {2.9158E-10, 3.0501E-09}

double	shrt_contig_transp_mdl_init [] = {1.3427E-08, 4.3168E-09}

double	non_contig_transp_mdl_init [] = {4.0475E-08, 4.0463E-09}

double	dgtog_res_mdl_init [] = {2.9786E-05, 2.4335E-04, 1.0845E-08}

double	blres_mdl_init [] = {1.0598E-05, 7.2741E-08}

double	alltoall_mdl_init [] = {1.0000E-06, 1.0000E-06, 5.0000E-10}

double	alltoallv_mdl_init [] = {2.7437E-06, 2.2416E-05, 1.0469E-08}

double	red_mdl_init [] = {6.2935E-07, 4.6276E-06, 9.2245E-10}

double	red_mdl_cst_init [] = {5.7302E-07, 4.7347E-06, 6.0191E-10}

double	allred_mdl_init [] = {8.4416E-07, 6.8651E-06, 3.5845E-08}

double	allred_mdl_cst_init [] = {-3.3754E-04, 2.1343E-04, 3.0801E-09}

double	bcast_mdl_init [] = {1.5045E-06, 1.4485E-05, 3.2876E-09}

double	spredist_mdl_init [] = {1.2744E-04, 1.0278E-03, 7.6837E-08}

double	csrred_mdl_init [] = {3.7005E-05, 1.1854E-04, 5.5165E-09}

double	csrred_mdl_cst_init [] = {-1.8323E-04, 1.3076E-04, 2.8732E-09}

double	upload_mdl_init []

double	download_mdl_init []

double	memcap = 0.5

int64_t	mem_size = 0

int	max_threads

int	instance_counter = 0

int64_t	mem_used [MAX_THREADS]

int64_t	tot_mem_used

int64_t	tot_mem_available = -1

std::list< mem_loc >	mem_stacks [MAX_THREADS]

void *	mst_buffer = 0

int64_t	mst_buffer_size = 0

int64_t	mst_buffer_used = 0

int64_t	mst_buffer_ptr = 0

std::list< mem_loc >	mst

char *	cpy_buffer [CPY_BUFFER_SIZE]

LinModel< 3 >	csrred_mdl (csrred_mdl_init,"csrred_mdl")

LinModel< 3 >	csrred_mdl_cst (csrred_mdl_cst_init,"csrred_mdl_cst")

LinModel< 3 >	spredist_mdl (spredist_mdl_init,"spredist_mdl")

Typedef Documentation

typedef bool CTF_int::TYPE1

Definition at line 6 of file ctf_ext.cxx.

typedef int CTF_int::TYPE2

Definition at line 7 of file ctf_ext.cxx.

typedef int64_t CTF_int::TYPE3

Definition at line 8 of file ctf_ext.cxx.

typedef float CTF_int::TYPE4

Definition at line 9 of file ctf_ext.cxx.

typedef double CTF_int::TYPE5

Definition at line 10 of file ctf_ext.cxx.

typedef std::complex<float> CTF_int::TYPE6

Definition at line 11 of file ctf_ext.cxx.

typedef std::complex<double> CTF_int::TYPE7

Definition at line 12 of file ctf_ext.cxx.

typedef int16_t CTF_int::TYPE8

Definition at line 13 of file ctf_ext.cxx.

typedef int8_t CTF_int::TYPE9

Definition at line 14 of file ctf_ext.cxx.

Enumeration Type Documentation

anonymous enum

Enumerator
SUCCESS
ERROR
NEGATIVE

Definition at line 97 of file common.h.

enum CTF_int::map_type

Enumerator
NOT_MAPPED
PHYSICAL_MAP
VIRTUAL_MAP

Definition at line 13 of file mapping.h.

enum CTF_int::TOPOLOGY

Enumerator
TOPOLOGY_GENERIC
TOPOLOGY_BGP
TOPOLOGY_BGQ
TOPOLOGY_8D
NO_TOPOLOGY

Definition at line 10 of file topology.h.

Function Documentation

template<typename dtype >

void CTF_int::abs_helper	(	tensor *	A,
		tensor *	B
	)

absolute value function

Parameters

[in] A tensor, param[in,out] B tensor (becomes absolute value of A)

Returns: None

Definition at line 17 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::abs(), ctf.core::dtype, and CTF_int::tensor::order.

template void CTF_int::abs_helper< bool >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< int16_t >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< int32_t >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< int64_t >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< int8_t >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::abs_helper< std::complex< double > >	(	tensor *	A,
		tensor *	B
	)

template void CTF_int::abs_helper< std::complex< float > >	(	tensor *	A,
		tensor *	B
	)

void CTF_int::add_sym_perm	(	std::vector< summation > &	perms,
		std::vector< int > &	signs,
		summation const &	new_perm,
		int	new_sign
	)

puts a summation map into a nice ordering according to preserved symmetries, and adds it if it is distinct

Parameters

[in,out]	perms	the permuted summation specifications
[in,out]	signs	sign of each summation
[in]	new_perm	summation signature
[in]	new_sign	alpha

Definition at line 549 of file symmetrization.cxx.

References CTF_int::summation::A, align_symmetric_indices(), CTF_int::summation::B, cdealloc(), CTF_int::summation::idx_A, CTF_int::summation::idx_B, inv_idx(), CTF_int::tensor::order, order_perm(), and CTF_int::tensor::sym.

void CTF_int::add_sym_perm	(	std::vector< contraction > &	perms,
		std::vector< int > &	signs,
		contraction const &	new_perm,
		int	new_sign
	)

puts a contraction map into a nice ordering according to preserved symmetries, and adds it if it is distinct

Parameters

[in,out]	perms	the permuted contraction specifications
[in,out]	signs	sign of each contraction
[in]	new_perm	contraction signature
[in]	new_sign	alpha

Definition at line 593 of file symmetrization.cxx.

References CTF_int::contraction::A, align_symmetric_indices(), CTF_int::contraction::B, CTF_int::contraction::C, cdealloc(), CTF_int::contraction::idx_A, CTF_int::contraction::idx_B, CTF_int::contraction::idx_C, inv_idx(), CTF_int::tensor::order, order_perm(), and CTF_int::tensor::sym.

Referenced by get_sym_perms().

template<typename dtype >

void CTF_int::all_helper	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

all function

Parameters

[in] A tensor, param[in] B tensor with bool values created, param[in] index of A, param[in] index of B

Returns: None

Definition at line 33 of file ctf_ext.cxx.

References ctf.core::a, and ctf.core::dtype.

template void CTF_int::all_helper< bool >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< double >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< float >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< int16_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< int32_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< int64_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< int8_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::all_helper< std::complex< double > >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

template void CTF_int::all_helper< std::complex< float > >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

void * CTF_int::alloc ( int64_t const len )

alloc abstraction

Parameters

[in] len number of bytes

Definition at line 365 of file memcontrol.cxx.

References alloc_ptr(), ASSERT, and SUCCESS.

Referenced by CTF_int::algstrct::alloc(), calc_drv_displs(), calc_sy_pfx(), calc_sy_pfx< 1 >(), compute_syoffs(), CTF_int::contraction::contraction(), conv_idx(), CTF_int::COO_Matrix::COO_Matrix(), CTF_int::CSR_Matrix::csr_add(), CTF_int::CSR_Matrix::CSR_Matrix(), CTF_int::algstrct::csr_reduce(), CTF::Bivar_Function< dtype_A, dtype_B, dtype_C >::csrmultcsr(), CTF::Bivar_Kernel< dtype_A, dtype_B, dtype_C, f, g >::csrmultcsr(), CTF::Bivar_Kernel< dtype_A, dtype_B, dtype_C, f, g >::csrmultcsr_old(), CTF_int::ctr_virt::ctr_virt(), cvrt_idx(), depad_tsr(), depermute_keys(), depin(), desymmetrize(), dgtog_reshuffle(), CTF_int::summation::estimate_time(), factorize(), CTF::Semiring< dtype, is_ord >::gen_csrmultcsr(), CTF_int::COO_Matrix::get_data(), get_full_intm(), get_grp_ptrs(), get_len_ordering(), CTF::Tensor< dtype >::get_local_data(), get_phys_topo(), CTF_int::algstrct::has_mul(), CTF::Idx_Tensor::Idx_Tensor(), CTF_int::tensor::init(), inv_idx(), CTF_int::LinModel< nparam >::LinModel(), CTF_int::Term::operator-(), CTF::Partition::operator=(), order_globally(), CTF_int::tensor::orient_subworld(), CTF_int::algstrct::pair_alloc(), CTF::Partition::Partition(), peel_torus(), CTF::Function_timer::print(), CTF::print_timers(), read_data_mpiio(), CTF_int::tensor::read_dense_from_file(), CTF::Tensor< dtype >::read_local(), CTF_int::tensor::read_local(), CTF_int::tensor::read_local_nnz(), CTF_int::tensor::redistribute(), CTF_int::ctr_virt::run(), CTF_int::spctr_replicate::run(), CTF_int::strp_tsr::run(), CTF_int::tspsum_virt::run(), CTF_int::tspsum_replicate::run(), CTF_int::spctr_2d_general::run(), CTF_int::tsum_virt::run(), CTF_int::spctr_virt::run(), CTF_int::tspsum_permute::run(), CTF_int::algstrct::safecopy(), CTF::Semiring< dtype, is_ord >::safemul(), CTF_int::scaling::scaling(), CTF_int::scl_virt::scl_virt(), CTF_int::tensor::self_reduce(), CTF_int::seq_tsr_ctr::seq_tsr_ctr(), CTF_int::seq_tsr_scl::seq_tsr_scl(), CTF_int::seq_tsr_spctr::seq_tsr_spctr(), CTF_int::seq_tsr_spsum::seq_tsr_spsum(), CTF_int::seq_tsr_sum::seq_tsr_sum(), serialize_sparse_tensor_data(), CTF_int::COO_Matrix::set_data(), CTF_int::tensor::set_name(), CTF_int::tensor::set_zero(), CTF_int::tensor::slice(), CTF::Tensor< dtype >::slice(), socopy(), spA_dnB_dnC_seq_ctr(), CTF_int::tensor::sparsify(), CTF_int::spctr_pin_keys::spctr_pin_keys(), CTF_int::spctr_virt::spctr_virt(), CTF_int::tensor::spmatricize(), spsfy_tsr(), CTF_int::summation::summation(), CTF::Matrix< dtype >::svd(), sym_seq_ctr_cust(), sym_seq_ctr_inr(), sym_seq_ctr_ref(), sym_seq_scl_cust(), sym_seq_scl_ref(), sym_seq_sum_cust(), sym_seq_sum_inr(), sym_seq_sum_ref(), symmetrize(), CTF_int::tensor::tensor(), CTF_int::topology::topology(), CTF_int::tspsum::tspsum(), CTF_int::tspsum_map::tspsum_map(), CTF_int::tspsum_permute::tspsum_permute(), CTF_int::tspsum_pin_keys::tspsum_pin_keys(), CTF_int::tspsum_virt::tspsum_virt(), CTF_int::tsum_virt::tsum_virt(), CTF_int::LinModel< nparam >::update(), and CTF_int::tensor::write_dense_to_file().

int CTF_int::alloc_ptr	(	int64_t const	len_,
		void **const	ptr
	)

alloc abstraction

Parameters

[in]	len_	number of bytes
[in,out]	ptr	pointer to set to new allocation address

Definition at line 320 of file memcontrol.cxx.

References ALIGN_BYTES, ASSERT, CTF_int::mem_loc::len, MAX, CTF_int::mem_loc::ptr, and SUCCESS.

template<typename dtype >

void CTF_int::any_helper	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

any function

Parameters

[in] A tensor, param[in] B tensor with bool values created, param[in] index of A, param[in] index of B

Returns: None

Definition at line 88 of file ctf_ext.cxx.

References ctf.core::a, and ctf.core::dtype.

template void CTF_int::any_helper< bool >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< double >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< float >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< int16_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< int32_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< int64_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< int8_t >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

Referenced by conv_type().

template void CTF_int::any_helper< std::complex< double > >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

template void CTF_int::any_helper< std::complex< float > >	(	tensor *	A,
		tensor *	B_bool,
		char const *	idx_A,
		char const *	idx_B
	)

void CTF_int::assign_keys	(	int	order,
		int64_t	size,
		int	nvirt,
		int const *	edge_len,
		int const *	sym,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_dim,
		int *	phase_rank,
		char const *	vdata,
		char *	vpairs,
		algstrct const *	sr
	)

assigns keys to an array of values

Parameters

[in]	order	tensor dimension
[in]	size	number of values
[in]	nvirt	total virtualization factor
[in]	edge_len	tensor edge lengths
[in]	sym	symmetries of tensor
[in]	phase	total phase of the tensor on virtualized processor grid
[in]	phys_phase	physical phase of the tensor
[in]	virt_dim	virtual phase in each dimension
[in]	phase_rank	physical phase rank multiplied by virtual phase
[in]	vdata	array of input values
[out]	vpairs	pairs of keys and inputted values
[in]	sr	algstrct defining data type of array

Definition at line 180 of file sparse_rw.cxx.

References ABORT, alloc_ptr(), ASSERT, cdealloc(), CTF_int::accumulatable::el_size, NS, CTF_int::algstrct::pair_size(), CTF_int::algstrct::set_pair(), TAU_FSTART, and TAU_FSTOP.

Referenced by read_loc_pairs().

char* CTF_int::bcast_step	(	int	edge_len,
		char *	A,
		bool	is_sparse_A,
		bool	move_A,
		algstrct const *	sr_A,
		int64_t	b_A,
		int64_t	s_A,
		char *	buf_A,
		CommData *	cdt_A,
		int64_t	ctr_sub_lda_A,
		int64_t	ctr_lda_A,
		int	nblk_A,
		int64_t const *	size_blk_A,
		int &	new_nblk_A,
		int64_t *&	new_size_blk_A,
		int64_t *	offsets_A,
		int	ib
	)

Definition at line 138 of file spctr_2d_general.cxx.

References alloc_ptr(), ASSERT, CTF_int::CommData::bcast(), cdealloc(), CTF_int::algstrct::copy(), CTF_int::spctr_2d_general::ctr_sub_lda_A, CTF_int::spctr_2d_general::edge_len, CTF_int::accumulatable::el_size, CTF_int::algstrct::mdtype(), mst_alloc_ptr(), CTF_int::CommData::np, CTF_int::CommData::rank, socopy(), and spcopy().

Referenced by CTF_int::spctr_2d_general::run().

void CTF_int::block_reshuffle	(	distribution const &	old_dist,
		distribution const &	new_dist,
		char *	tsr_data,
		char *&	tsr_cyclic_data,
		algstrct const *	sr,
		CommData	glb_comm
	)

Reshuffle elements by block given the global phases stay the same.

Parameters

[in]	old_dist	starting data distrubtion
[in]	new_dist	target data distrubtion
[in]	tsr_data	starting data buffer
[out]	tsr_cyclic_data	target data buffer
[in]	sr	algstrct defining data
[in]	glb_comm	communicator on which to redistribute

Definition at line 454 of file redist.cxx.

References CTF_int::algstrct::addid(), CTF_int::algstrct::alloc(), alloc_ptr(), cdealloc(), CTF_int::CommData::cm, CTF_int::algstrct::copy(), DPRINTF, CTF_int::accumulatable::el_size, CTF_int::algstrct::mdtype(), CTF_int::LinModel< nparam >::observe(), CTF_int::distribution::order, CTF_int::distribution::pe_lda, CTF_int::distribution::perank, CTF_int::distribution::phase, CTF_int::distribution::phys_phase, CTF_int::CommData::rank, CTF_int::algstrct::set(), CTF_int::LinModel< nparam >::should_observe(), CTF_int::distribution::size, TAU_FSTART, TAU_FSTOP, and CTF_int::distribution::virt_phase.

Referenced by CTF_int::tensor::redistribute().

double CTF_int::blres_est_time	(	int64_t	tot_sz,
		int	nv0,
		int	nv1
	)

estimates execution time, given this processor sends a receives tot_sz across np procs

Parameters

[in]	tot_sz	amount of data sent/recved
[in]	nv0	starting number of blocks
[in]	nv1	ending number of blocks

Definition at line 449 of file redist.cxx.

References CTF_int::LinModel< nparam >::est_time().

Referenced by CTF_int::tensor::est_redist_time().

void CTF_int::bucket_by_pe	(	int	order,
		int64_t	num_pair,
		int64_t	np,
		int const *	phys_phase,
		int const *	virt_phase,
		int const *	bucket_lda,
		int const *	edge_len,
		ConstPairIterator	mapped_data,
		int64_t *	bucket_counts,
		int64_t *	bucket_off,
		PairIterator	bucket_data,
		algstrct const *	sr
	)

buckets key-value pairs by processor according to distribution

Parameters

[in]	order	number of tensor dims
[in]	num_pair	numbers of values being written
[in]	np	number of processor buckets
[in]	phys_phase	physical distribution phase
[in]	virt_phase	factor of phase due to local blocking
[in]	bucket_lda	iterator hop along each bucket dim
[in]	edge_len	padded edge lengths of tensor
[in]	mapped_data	set of sparse key-value pairs
[out]	bucket_counts	how many keys belong to each processor
[out]	bucket_off	prefix sum of bucket_counts
[out]	bucket_data	mapped_data reordered by bucket
[in]	sr	algstrct context defining values

Definition at line 432 of file sparse_rw.cxx.

References alloc_ptr(), ASSERT, cdealloc(), CTF_int::ConstPairIterator::k(), ctf.core::np(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write().

Referenced by wr_pairs_layout().

int64_t * CTF_int::bucket_by_virt	(	int	order,
		int	num_virt,
		int64_t	num_pair,
		int const *	phys_phase,
		int const *	virt_phase,
		int const *	edge_len,
		ConstPairIterator	mapped_data,
		PairIterator	bucket_data,
		algstrct const *	sr
	)

buckets key value pairs by block/virtual-processor

Parameters

[in]	order	number of tensor dims
[in]	num_virt	number of local blocks
[in]	num_pair	numbers of values being written
[in]	phys_phase	physical distribution phase
[in]	virt_phase	factor of phase due to local blocking
[in]	edge_len	padded edge lengths of tensor
[in]	mapped_data	set of sparse key-value pairs
[out]	bucket_data	mapped_data reordered by bucket
[in]	sr	algstrct context defining values

Definition at line 539 of file sparse_rw.cxx.

References alloc_ptr(), ASSERT, cdealloc(), CTF_int::ConstPairIterator::k(), CTF_int::PairIterator::sort(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write().

Referenced by wr_pairs_layout().

template<int idim>

int64_t CTF_int::calc_cnt	(	int const *	sym,
		int const *	rep_phase,
		int const *	sphase,
		int const *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

computes the cardinality of the set of elements of a tensor of order idim+1 that are owned by processor index gidx_off in a distribution with dimensions sphase

Definition at line 23 of file dgtog_calc_cnt.cxx.

References cdealloc(), get_loc(), and NS.

Referenced by calc_drv_cnts< 0 >(), and calc_sy_pfx().

template<>

int64_t CTF_int::calc_cnt< 0 >	(	int const *	sym,
		int const *	rep_phase,
		int const *	sphase,
		int const *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

Definition at line 44 of file dgtog_calc_cnt.cxx.

References get_loc(), and NS.

Referenced by calc_sy_pfx< 1 >().

void CTF_int::calc_cnt_displs	(	int const *	sym,
		distribution const &	old_dist,
		distribution const &	new_dist,
		int	new_nvirt,
		int	np,
		int const *	old_virt_edge_len,
		int const *	new_virt_lda,
		int64_t *	send_counts,
		int64_t *	recv_counts,
		int64_t *	send_displs,
		int64_t *	recv_displs,
		CommData	ord_glb_comm,
		int	idx_lyr,
		int const	bucket_offset
	)

assigns keys to an array of values

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	old_dist	starting data distrubtion
[in]	new_dist	target data distrubtion
[in]	new_nvirt	new total virtualization factor
[in]	np	number of processors
[in]	old_virt_edge_len	old edge lengths of blocks
[in]	new_virt_lda	prefix sum of new_dist.virt_phase
[out]	send_counts	outgoing counts of pairs by pe
[out]	recv_counts	incoming counts of pairs by pe
[out]	send_displs	outgoing displs of pairs by pe
[out]	recv_displs	incoming displs of pairs by pe
[in]	ord_glb_comm	the global communicator
[in]	idx_lyr	starting processor layer (2.5D)
[in]	bucket_offset	offsets for target index for each dimension

Definition at line 170 of file redist.cxx.

References alloc_ptr(), ASSERT, blres_mdl, blres_mdl_init, cdealloc(), CTF_int::CommData::cm, ctf.core::dim, MAX, MIN, mst_alloc_ptr(), ctf.core::np(), NS, CTF_int::distribution::order, CTF_int::distribution::pad_edge_len, CTF_int::distribution::padding, CTF_int::distribution::perank, CTF_int::distribution::phase, CTF_int::distribution::phys_phase, SY, sy_packed_size(), and CTF_int::distribution::virt_phase.

Referenced by cyclic_reshuffle(), and glb_cyclic_reshuffle().

template<int idim>

void CTF_int::calc_cnt_from_rep_cnt	(	int const *	rep_phase,
		int const	pe_offset,
		int const	bucket_offset,
		int64_t const *	old_counts,
		int64_t *	counts,
		int	bucket_off,
		int	pe_off,
		int	dir
	)

Definition at line 155 of file dgtog_calc_cnt.cxx.

References calc_cnt_from_rep_cnt< 0 >().

Referenced by dgtog_reshuffle().

template<>

void CTF_int::calc_cnt_from_rep_cnt< 0 >	(	int const *	rep_phase,
		int const	pe_offset,
		int const	bucket_offset,
		int64_t const *	old_counts,
		int64_t *	counts,
		int	bucket_off,
		int	pe_off,
		int	dir
	)

Definition at line 174 of file dgtog_calc_cnt.cxx.

Referenced by calc_cnt_from_rep_cnt().

void CTF_int::calc_dim	(	int	order,
		int64_t	size,
		int const *	edge_len,
		mapping const *	edge_map,
		int64_t *	vrt_sz,
		int *	vrt_edge_len,
		int *	blk_edge_len
	)

calculate the block-sizes of a tensor

Parameters

[in]	order	number of dimensions of this tensor
[in]	size	is the size of the local tensor stored
[in]	edge_len	edge lengths of global tensor
[in]	edge_map	mapping of each dimension
[out]	vrt_sz	size of virtual block
[out]	vrt_edge_len	edge lengths of virtual block
[out]	blk_edge_len	edge lengths of local block

Definition at line 143 of file distribution.cxx.

References CTF_int::mapping::child, CTF_int::mapping::has_child, CTF_int::mapping::np, CTF_int::distribution::order, PHYSICAL_MAP, CTF_int::distribution::size, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by CTF_int::tensor::est_time_unfold(), CTF_int::summation::estimate_time(), CTF_int::scaling::execute(), CTF_int::tensor::fold(), get_len_ordering(), and CTF_int::tensor::unfold().

template<int idim>

void CTF_int::calc_drv_cnts	(	int	order,
		int const *	sym,
		int64_t *	counts,
		int const *	rep_phase,
		int const *	rep_phase_lda,
		int const *	sphase,
		int const *	phys_phase,
		int *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

Definition at line 120 of file dgtog_calc_cnt.cxx.

Referenced by calc_drv_displs().

template<>

void CTF_int::calc_drv_cnts< 0 >	(	int	order,
		int const *	sym,
		int64_t *	counts,
		int const *	rep_phase,
		int const *	rep_phase_lda,
		int const *	sphase,
		int const *	phys_phase,
		int *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

Definition at line 138 of file dgtog_calc_cnt.cxx.

References calc_cnt(), and SWITCH_ORD_CALL_RET.

void CTF_int::calc_drv_displs	(	int const *	sym,
		int const *	edge_len,
		distribution const &	old_dist,
		distribution const &	new_dist,
		int64_t *	counts,
		int	idx_lyr
	)

Definition at line 220 of file dgtog_calc_cnt.cxx.

References alloc(), calc_drv_cnts(), cdealloc(), lcm(), SWITCH_ORD_CALL, TAU_FSTART, and TAU_FSTOP.

Referenced by dgtog_reshuffle().

void CTF_int::calc_fold_lnmk	(	tensor const *	A,
		tensor const *	B,
		tensor const *	C,
		int const *	idx_A,
		int const *	idx_B,
		int const *	idx_C,
		int const *	ordering_A,
		int const *	ordering_B,
		iparam *	inner_prm
	)

calculate the dimensions of the matrix the contraction gets reduced to (A, B, and C may be permuted)

Parameters

[in]	A	tensor 1
[in]	B	tensor 2
[in]	C	tensor 3
[in]	idx_A	indices of tensor 1
[in]	idx_B	indices of tensor 2
[in]	idx_C	indices of tensor 3
[in]	ordering_A	the dimensional-ordering of the inner mapping of A
[in]	ordering_B	the dimensional-ordering of the inner mapping of B
[out]	inner_prm	parameters includng l(number of matrix mutlplications),n,m,k

Definition at line 200 of file contraction.cxx.

References alloc_ptr(), cdealloc(), inv_idx(), CTF_int::iparam::k, CTF_int::iparam::l, CTF_int::iparam::m, CTF_int::iparam::n, NS, CTF_int::tensor::order, CTF_int::tensor::pad_edge_len, and CTF_int::iparam::sz_C.

Referenced by get_len_ordering().

void CTF_int::calc_idx_arr	(	int	order,
		int const *	lens,
		int const *	sym,
		int64_t	idx,
		int *	idx_arr
	)

Definition at line 72 of file util.cxx.

References ASSERT, ctf.core::dim, NS, and packed_size().

Referenced by get_choice(), glb_ord_pup(), pad_cyclic_pup_virt_buff(), scal_diag(), and zero_padding().

template<int idim>

int64_t * CTF_int::calc_sy_pfx	(	int const *	sym,
		int const *	rep_phase,
		int const *	sphase,
		int const *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

computes the cardinality of the sets of elements of a tensor of order idim+1 for different values of the idim'th tensor dimension

Definition at line 55 of file dgtog_calc_cnt.cxx.

References alloc(), calc_cnt(), cdealloc(), get_glb(), get_loc(), NS, and SY.

template<>

int64_t* CTF_int::calc_sy_pfx< 1 >	(	int const *	sym,
		int const *	rep_phase,
		int const *	sphase,
		int const *	gidx_off,
		int const *	edge_len,
		int const *	loc_edge_len
	)

Definition at line 97 of file dgtog_calc_cnt.cxx.

References alloc(), calc_cnt< 0 >(), get_glb(), get_loc(), NS, and SY.

int CTF_int::can_block_reshuffle	(	int	order,
		int const *	old_phase,
		mapping const *	map
	)

determines if tensor can be permuted by block

Parameters

[in]	order	dimension of tensor
[in]	old_phase	old cyclic phases in each dimension
[in]	map	new mapping for each edge length

Returns: 1 if block reshuffle allowed, 0 if not

Definition at line 618 of file redist.cxx.

References CTF_int::mapping::calc_phase().

Referenced by CTF_int::tensor::est_redist_time(), CTF_int::tensor::get_redist_mem(), CTF_int::summation::is_equal(), and CTF_int::tensor::redistribute().

int CTF_int::can_morph	(	topology const *	topo_keep,
		topology const *	topo_change
	)

determines if two topologies are compatible with each other

Parameters

topo_keep	topology to keep (larger dimension)
topo_change	topology to change (smaller dimension)

Returns: true if its possible to change

Definition at line 683 of file topology.cxx.

References CTF_int::topology::dim_comm, CTF_int::topology::lda, CTF_int::CommData::np, and CTF_int::topology::order.

Referenced by get_len_ordering().

double CTF_int::cddot	(	int	n,
		const double *	dX,
		int	incX,
		const double *	dY,
		int	incY
	)

Definition at line 60 of file model.cxx.

References CTF_BLAS::DDOT().

int CTF_int::cdealloc ( void * ptr )

free abstraction

Parameters

[in,out] ptr pointer to set to address to free

Definition at line 480 of file memcontrol.cxx.

References ABORT, cdealloc(), ERROR, max_threads, mst_free(), NEGATIVE, and SUCCESS.

int CTF_int::cdealloc	(	void *	ptr,
		int const	tid
	)

free abstraction

Parameters

[in,out]	ptr	pointer to set to address to free
[in]	tid	thread id from whose stack pointer needs to be freed

Definition at line 410 of file memcontrol.cxx.

References ctf.core::it, mst_free(), NEGATIVE, and SUCCESS.

Referenced by cdealloc(), and cdealloc_cond().

int CTF_int::cdealloc_cond ( void * ptr )

free abstraction (conditional (no error if not found))

Parameters

[in,out] ptr pointer to set to address to free

Definition at line 448 of file memcontrol.cxx.

References cdealloc(), max_threads, NEGATIVE, and SUCCESS.

void CTF_int::cdgelsd	(	int	m,
		int	n,
		int	k,
		double const *	A,
		int	lda_A,
		double *	B,
		int	lda_B,
		double *	S,
		double	cond,
		int *	rank,
		double *	work,
		int	lwork,
		int *	iwork,
		int *	info
	)

Definition at line 102 of file model.cxx.

References CTF_LAPACK::cdgelsd().

Referenced by CTF_int::LinModel< nparam >::update().

void CTF_int::cdgeqrf	(	int const	M,
		int const	N,
		double *	A,
		int const	LDA,
		double *	TAU2,
		double *	WORK,
		int const	LWORK,
		int *	INFO
	)

Definition at line 67 of file model.cxx.

References CTF_LAPACK::cdgeqrf().

Referenced by CTF_int::LinModel< nparam >::update().

void CTF_int::cdormqr	(	char	SIDE,
		char	TRANS,
		int	M,
		int	N,
		int	K,
		double const *	A,
		int	LDA,
		double const *	TAU2,
		double *	C,
		int	LDC,
		double *	WORK,
		int	LWORK,
		int *	INFO
	)

Definition at line 80 of file model.cxx.

References CTF_LAPACK::cdormqr().

Referenced by CTF_int::LinModel< nparam >::update().

template<typename dtype , dtype(*)(dtype) abs>

void CTF_int::char_abs	(	char const *	a,
		char *	b
	)

Definition at line 130 of file set.h.

References ctf.core::abs(), and ctf.core::dtype.

int64_t CTF_int::chchoose	(	int64_t	n,
		int64_t	k
	)

Definition at line 305 of file util.cxx.

References fact().

Referenced by coalesce_bwd().

int CTF_int::check_self_mapping	(	tensor const *	tsr,
		int const *	idx_map
	)

checks mapping in preparation for tensors scale, summ or contract

Parameters

[in]	tsr	handle to tensor
[in]	idx_map	is the mapping of tensor to global indices

Returns: whether the self mapping is consistent

Definition at line 332 of file mapping.cxx.

References alloc_ptr(), CTF_int::mapping::calc_phase(), cdealloc(), CTF_int::mapping::cdt, CTF_int::mapping::child, DPRINTF, CTF_int::tensor::edge_map, CTF_int::mapping::has_child, CTF_int::tensor::order, PHYSICAL_MAP, and CTF_int::mapping::type.

Referenced by CTF_int::scaling::execute(), get_len_ordering(), CTF_int::summation::is_equal(), and CTF_int::tensor::set_distribution().

int64_t CTF_int::choose	(	int64_t	n,
		int64_t	k
	)

Definition at line 285 of file util.cxx.

References fact().

Referenced by coalesce_bwd(), and get_len_ordering().

void CTF_int::cmp_sym_perms	(	int	ndim,
		int const *	sym,
		int *	nperm,
		int **	perm,
		double *	sign
	)

finds all permutations of a tensor according to a symmetry

Parameters

[in]	ndim	dimension of tensor
[in]	sym	symmetry specification of tensor
[out]	nperm	number of symmeitrc permutations to do
[out]	perm	the permutation
[out]	sign	sign of each permutation

Definition at line 400 of file symmetrization.cxx.

References alloc_ptr(), AS, ASSERT, ctf.core::np(), and NS.

void CTF_int::coalesce_bwd	(	int	el_size,
		char *	B,
		char const *	B_aux,
		int	k,
		int	n,
		int	kb
	)

inline

we receive a contiguous buffer kb-by-n B and (k-kb)-by-n B_aux which is the block below. To get a k-by-n buffer, we need to combine this buffer with our original block. Since we are working with column-major ordering we need to interleave the blocks. Thats what this function does.

Parameters

[in]	el_size	element size
[in,out]	B	the buffer to coalesce into
[in]	B_aux	the second buffer to coalesce from
[in]	k	the total number of rows
[in]	n	the number of columns
[in]	kb	the number of rows in a B originally

Definition at line 398 of file util.h.

References ctf.core::a, ctf.core::b, chchoose(), choose(), fact(), get_choice(), permute(), permute_target(), socopy(), and spcopy().

int CTF_int::comp_dim_map	(	mapping const *	map_A,
		mapping const *	map_B
	)

compares two mappings

Parameters

map_A	first map
map_B	second map return true if mapping is exactly the same, false otherwise

Definition at line 143 of file mapping.cxx.

References ASSERT, CTF_int::mapping::cdt, CTF_int::mapping::child, DEBUG_PRINTF, CTF_int::mapping::has_child, NOT_MAPPED, CTF_int::mapping::np, PHYSICAL_MAP, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by CTF_int::tensor::align(), ctr_2d_gen_build(), get_len_ordering(), and CTF_int::summation::is_equal().

template<int nparam>

bool CTF_int::comp_time_param	(	const time_param< nparam > &	a,
		const time_param< nparam > &	b
	)

Definition at line 114 of file model.cxx.

References CTF_int::time_param< nparam >::p.

int ** CTF_int::compute_bucket_offsets	(	distribution const &	old_dist,
		distribution const &	new_dist,
		int const *	len,
		int const *	old_phys_edge_len,
		int const *	old_virt_lda,
		int const *	old_offsets,
		int const	old_permutation,
		int const *	new_phys_edge_len,
		int const *	new_virt_lda,
		int	forward,
		int	old_virt_np,
		int	new_virt_np,
		int const *	old_virt_edge_len
	)

computes offsets for redistribution targets along each edge length

Parameters

[in]	old_dist	starting data distrubtion
[in]	new_dist	target data distrubtion
[in]	len	unpadded edge lengths
[in]	old_phys_edge_len	total edge lengths of old local tensor chunk
[in]	old_virt_lda	prefix sum of old_dist.virt_phase
[in]	old_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	old_permutation	permutation array for each edge length (no perm if NULL)
[in]	new_phys_edge_len	total edge lengths of new local tensor chunk
[in]	new_virt_lda	prefix sum of new_dist.virt_phase
[in]	forward	1 for sending 0 for receiving
[in]	old_virt_np	number of blocks per processor in old_dist
[in]	new_virt_np	number of blocks per processor in new_dist
[in]	old_virt_edge_len	edge lengths of each block in old_dist

Returns: 2D array with dims [order][old_phys_edge_len[i]] with bucket offsets for each edge length

Definition at line 111 of file redist.cxx.

References alloc_ptr(), ctf.core::dim, MAX, CTF_int::distribution::order, CTF_int::distribution::pe_lda, CTF_int::distribution::perank, CTF_int::distribution::phase, CTF_int::distribution::phys_phase, TAU_FSTART, TAU_FSTOP, and CTF_int::distribution::virt_phase.

Referenced by cyclic_reshuffle(), and glb_cyclic_reshuffle().

void CTF_int::compute_syoff	(	int	r,
		int	len,
		algstrct const *	sr,
		int const *	edge_len,
		int const *	sym,
		uint64_t *	offsets
	)

Definition at line 301 of file sym_seq_ctr.cxx.

References CTF_int::accumulatable::el_size, NS, and sy_packed_size().

Referenced by compute_syoffs().

void CTF_int::compute_syoffs	(	algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		int	tot_order,
		int const *	rev_idx_map,
		uint64_t **&	offsets_A,
		uint64_t **&	offsets_B
	)

Definition at line 193 of file sym_seq_sum.cxx.

References alloc(), compute_syoff(), compute_syoffs(), TAU_FSTART, and TAU_FSTOP.

void CTF_int::compute_syoffs	(	algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		int	tot_order,
		int const *	rev_idx_map,
		uint64_t **&	offsets_A,
		uint64_t **&	offsets_B,
		uint64_t **&	offsets_C
	)

Definition at line 332 of file sym_seq_ctr.cxx.

References alloc(), compute_syoff(), TAU_FSTART, and TAU_FSTOP.

Referenced by compute_syoffs(), spA_dnB_dnC_seq_ctr(), sym_seq_ctr_cust(), sym_seq_ctr_ref(), and sym_seq_sum_ref().

template<typename dtype >

void CTF_int::conj_helper	(	tensor *	A,
		tensor *	B
	)

Definition at line 42 of file ctf_ext.cxx.

References ctf.core::a, and CTF_int::tensor::order.

template void CTF_int::conj_helper< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::conj_helper< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

std::vector< Term* > CTF_int::contract_down_terms	(	algstrct *	sr,
		char *	tscale,
		std::vector< Term * >	operands,
		std::vector< char >	out_inds,
		int	terms_to_leave,
		bool	est_time = `false`,
		double *	cost = `NULL`
	)

Definition at line 563 of file term.cxx.

References CTF::Idx_Tensor::clone(), CTF_int::Contract_Term::clone(), det_uniq_inds(), CTF_int::Term::estimate_time(), CTF_int::contraction::estimate_time(), CTF_int::Term::execute(), CTF_int::contraction::execute(), get_full_intm(), CTF::Idx_Tensor::idx_map, CTF_int::algstrct::mulid(), CTF::Idx_Tensor::parent, CTF_int::algstrct::safecopy(), CTF_int::algstrct::safemul(), and CTF_int::Term::scale.

Referenced by CTF_int::Contract_Term::estimate_time(), and CTF_int::Contract_Term::execute().

std::list< mem_transfer > CTF_int::contract_mst ( )

gets rid of empty space on the stack

Definition at line 125 of file memcontrol.cxx.

References CPY_BUFFER_SIZE, ctf.core::it, MIN, mst, mst_alloc(), CTF_int::mem_transfer::new_ptr, CTF_int::mem_transfer::old_ptr, TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::summation::estimate_time(), and get_len_ordering().

template<typename type >

int CTF_int::conv_idx	(	int	order,
		type const *	cidx,
		int **	iidx
	)

Definition at line 50 of file common.cxx.

References alloc().

Referenced by CTF_int::contraction::contraction(), CTF_int::scaling::scaling(), CTF_int::tensor::set_distribution(), and CTF_int::summation::summation().

template<typename type >

int CTF_int::conv_idx	(	int	order_A,
		type const *	cidx_A,
		int **	iidx_A,
		int	order_B,
		type const *	cidx_B,
		int **	iidx_B
	)

Definition at line 76 of file common.cxx.

References alloc(), and conv_idx().

template<typename type >

int CTF_int::conv_idx	(	int	order_A,
		type const *	cidx_A,
		int **	iidx_A,
		int	order_B,
		type const *	cidx_B,
		int **	iidx_B,
		int	order_C,
		type const *	cidx_C,
		int **	iidx_C
	)

Definition at line 114 of file common.cxx.

References alloc(), conv_idx< char >(), and conv_idx< int >().

Referenced by conv_idx().

template int CTF_int::conv_idx< char >	(	int	,
		char const *	,
		int **
	)

template int CTF_int::conv_idx< char >	(	int	,
		char const *	,
		int **	,
		int	,
		char const *	,
		int **
	)

template int CTF_int::conv_idx< char >	(	int	,
		char const *	,
		int **	,
		int	,
		char const *	,
		int **	,
		int	,
		char const *	,
		int **
	)

Referenced by conv_idx().

template int CTF_int::conv_idx< int >	(	int	,
		int const *	,
		int **
	)

Referenced by CTF_int::summation::estimate_time(), and get_len_ordering().

template int CTF_int::conv_idx< int >	(	int	,
		int const *	,
		int **	,
		int	,
		int const *	,
		int **
	)

template int CTF_int::conv_idx< int >	(	int	,
		int const *	,
		int **	,
		int	,
		int const *	,
		int **	,
		int	,
		int const *	,
		int **
	)

Referenced by conv_idx().

void CTF_int::conv_type	(	int	type_idx1,
		int	type_idx2,
		tensor *	A,
		tensor *	B
	)

convert tensor from one type to another

Parameters

[in]	type_idx1	index of first ype
[in]	type_idx2	index of second ype
[in]	A	tensor to convert
[in]	B	tensor to convert to

Definition at line 327 of file ctf_ext.cxx.

References abs_helper< bool >(), abs_helper< double >(), abs_helper< float >(), abs_helper< int16_t >(), abs_helper< int32_t >(), abs_helper< int64_t >(), abs_helper< int8_t >(), all_helper< bool >(), all_helper< double >(), all_helper< float >(), all_helper< int16_t >(), all_helper< int32_t >(), all_helper< int64_t >(), all_helper< int8_t >(), any_helper< bool >(), any_helper< double >(), any_helper< float >(), any_helper< int16_t >(), any_helper< int32_t >(), any_helper< int64_t >(), any_helper< int8_t >(), conj_helper< double >(), conj_helper< float >(), get_imag< double >(), get_imag< float >(), get_real< double >(), get_real< float >(), pow_helper< bool >(), pow_helper< double >(), pow_helper< float >(), pow_helper< int16_t >(), pow_helper< int32_t >(), pow_helper< int64_t >(), pow_helper< int8_t >(), set_imag< double >(), set_imag< float >(), set_real< double >(), set_real< float >(), and SWITCH_TYPE.

void CTF_int::copy_mapping	(	int	order,
		mapping const *	mapping_A,
		mapping *	mapping_B
	)

copies mapping A to B

Parameters

[in]	order	number of dimensions
[in]	mapping_A	mapping to copy from
[in,out]	mapping_B	mapping to copy to

Definition at line 190 of file mapping.cxx.

References CTF_int::mapping::clear(), and CTF_int::mapping::mapping().

Referenced by CTF_int::tensor::align(), copy_mapping(), CTF_int::tensor::copy_tensor_data(), desymmetrize(), CTF_int::summation::estimate_time(), CTF_int::scaling::execute(), get_len_ordering(), CTF_int::summation::is_equal(), CTF_int::tensor::pull_alias(), and symmetrize().

int CTF_int::copy_mapping	(	int	order_A,
		int	order_B,
		int const *	idx_A,
		mapping const *	mapping_A,
		int const *	idx_B,
		mapping *	mapping_B,
		int	make_virt = `1`
	)

copies mapping A to B

Parameters

[in]	order_A	number of dimensions in A
[in]	order_B	number of dimensions in B
[in]	idx_A	index mapping of A
[in]	mapping_A	mapping to copy from
[in]	idx_B	index mapping of B
[in,out]	mapping_B	mapping to copy to
[in]	make_virt	makes virtual

Definition at line 210 of file mapping.cxx.

References ASSERT, cdealloc(), CTF_int::mapping::clear(), copy_mapping(), CTF_int::mapping::has_child, inv_idx(), CTF_int::mapping::np, SUCCESS, CTF_int::mapping::type, and VIRTUAL_MAP.

int CTF_int::ctr_2d_gen_build	(	int	is_used,
		CommData	global_comm,
		int	i,
		int *	virt_dim,
		int &	cg_edge_len,
		int &	total_iter,
		tensor *	A,
		int	i_A,
		CommData *&	cg_cdt_A,
		int64_t &	cg_ctr_lda_A,
		int64_t &	cg_ctr_sub_lda_A,
		bool &	cg_move_A,
		int *	blk_len_A,
		int64_t &	blk_sz_A,
		int const *	virt_blk_len_A,
		int &	load_phase_A,
		tensor *	B,
		int	i_B,
		CommData *&	cg_cdt_B,
		int64_t &	cg_ctr_lda_B,
		int64_t &	cg_ctr_sub_lda_B,
		bool &	cg_move_B,
		int *	blk_len_B,
		int64_t &	blk_sz_B,
		int const *	virt_blk_len_B,
		int &	load_phase_B,
		tensor *	C,
		int	i_C,
		CommData *&	cg_cdt_C,
		int64_t &	cg_ctr_lda_C,
		int64_t &	cg_ctr_sub_lda_C,
		bool &	cg_move_C,
		int *	blk_len_C,
		int64_t &	blk_sz_C,
		int const *	virt_blk_len_C,
		int &	load_phase_C
	)

sets up a ctr_2d_general (2D SUMMA) level where A is not communicated function will be called with A/B/C permuted depending on desired alg

Parameters

[in]	is_used	whether this ctr will actually be run
[in]	global_comm	comm for this CTF instance
[in]	i	index in the total index map currently worked on
[in,out]	virt_dim	virtual processor grid lengths
[out]	cg_edge_len	edge lengths of ctr_2d_gen object to set
[in,out]	total_iter	the total number of ctr_2d_gen iterations
[in]	A	A tensor
[in]	i_A	the index in A to which index i corresponds
[out]	cg_cdt_A	the communicator for A to be set for ctr_2d_gen
[out]	cg_ctr_lda_A	parameter of ctr_2d_gen corresponding to upper lda for lda_cpy
[out]	cg_ctr_sub_lda_A	parameter of ctr_2d_gen corresponding to lower lda for lda_cpy
[out]	cg_move_A	tells ctr_2d_gen whether A should be communicated
[in,out]	blk_len_A	lengths of local A piece after this ctr_2d_gen level
[in,out]	blk_sz_A	size of local A piece after this ctr_2d_gen level
[in]	virt_blk_edge_len_A	edge lengths of virtual blocks of A
[in]	load_phase_A	tells the offloader how often A buffer changes for ctr_2d_gen

... the other parameters are specified the same as for _A but this time for _B and _C

Definition at line 12 of file ctr_2d_general.cxx.

References ASSERT, CTF_int::mapping::calc_phase(), CTF_int::mapping::cdt, CTF_int::mapping::child, comp_dim_map(), CTF_int::topology::dim_comm, CTF_int::tensor::edge_map, CTF_int::mapping::has_child, lcm(), MAX, CTF_int::mapping::np, CTF_int::tensor::order, PHYSICAL_MAP, CTF_int::tensor::topo, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by CTF_int::contraction::contraction(), and get_len_ordering().

void CTF_int::cvrt_idx	(	int	order,
		int const *	lens,
		int64_t	idx,
		int *	idx_arr
	)

Definition at line 533 of file common.cxx.

Referenced by CTF::Tensor< dtype >::slice(), and wr_pairs_layout().

void CTF_int::cvrt_idx	(	int	order,
		int const *	lens,
		int64_t	idx,
		int **	idx_arr
	)

Definition at line 545 of file common.cxx.

References alloc(), and cvrt_idx().

void CTF_int::cvrt_idx	(	int	order,
		int const *	lens,
		int const *	idx_arr,
		int64_t *	idx
	)

Definition at line 553 of file common.cxx.

References MPI_CTF_DOUBLE_COMPLEX.

Referenced by cvrt_idx().

void CTF_int::cyclic_reshuffle	(	int const *	sym,
		distribution const &	old_dist,
		int const *	old_offsets,
		int const	old_permutation,
		distribution const &	new_dist,
		int const *	new_offsets,
		int const	new_permutation,
		char **	tsr_data,
		char **	tsr_cyclic_data,
		algstrct const *	sr,
		CommData	ord_glb_comm,
		bool	reuse_buffers,
		char const *	alpha,
		char const *	beta
	)

Goes from any set of phases to any new set of phases.

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	old_dist	starting data distrubtion
[in]	old_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	old_permutation	permutation array for each edge length (no perm if NULL)
[in]	new_dist	target data distrubtion
[in]	new_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	new_permutation	permutation array for each edge length (no perm if NULL)
[in]	tsr_data	starting data buffer
[out]	tsr_cyclic_data	target data buffer
[in]	sr	algstrct defining data
[in]	ord_glb_comm	communicator on which to redistribute
[in]	reuse_buffers	if 1: ptr_tsr_cyclic_data is allocated dynamically and ptr_tsr_data is overwritten with intermediate data if 0: ptr_tsr_cyclic_data is preallocated and can be scaled by beta, however, more memory is used for temp buffers
[in]	alpha	scaling tensor for new data
[in]	beta	scaling tensor for original data

Definition at line 477 of file cyclic_reshuffle.cxx.

Referenced by CTF_int::tensor::add_from_subworld(), CTF_int::tensor::add_to_subworld(), and glb_cyclic_reshuffle().

template<typename dtype >

void CTF_int::def_coo_to_csr	(	int64_t	nz,
		int	nrow,
		dtype *	csr_vs,
		int *	csr_ja,
		int *	csr_ia,
		dtype const *	coo_vs,
		int const *	coo_rs,
		int const *	coo_cs
	)

Definition at line 100 of file set.h.

Referenced by CTF::Set< dtype, is_ord >::coo_to_csr().

template<typename dtype >

void CTF_int::def_csr_to_coo	(	int64_t	nz,
		int	nrow,
		dtype const *	csr_vs,
		int const *	csr_ja,
		int const *	csr_ia,
		dtype *	coo_vs,
		int *	coo_rs,
		int *	coo_cs
	)

Definition at line 105 of file set.h.

Referenced by CTF::Set< dtype, is_ord >::csr_to_coo().

template<typename dtype , bool is_ord>

std::enable_if<is_ord, dtype>::type CTF_int::default_abs ( dtype a )

inline

Definition at line 116 of file set.h.

References ctf.core::a, ctf.core::b, and ctf.core::dtype.

template<typename dtype , bool is_ord>

std::enable_if<!is_ord, dtype>::type CTF_int::default_abs ( dtype a )

inline

Definition at line 123 of file set.h.

References ctf.core::a.

template<typename dtype >

dtype CTF_int::default_add	(	dtype	a,
		dtype	b
	)

Definition at line 6 of file monoid.h.

References ctf.core::b.

template<typename dtype >

dtype CTF_int::default_addinv ( dtype a )

Definition at line 110 of file set.h.

References ctf.core::a.

template<typename dtype >

void CTF_int::default_axpy	(	int	n,
		dtype	alpha,
		dtype const *	X,
		int	incX,
		dtype *	Y,
		int	incY
	)

Definition at line 19 of file semiring.h.

References default_axpy< double >(), and default_axpy< float >().

template<>

void CTF_int::default_axpy< double >	(	int	n,
		double	alpha,
		double const *	X,
		int	incX,
		double *	Y,
		int	incY
	)

Definition at line 139 of file semiring.cxx.

References CTF_BLAS::DAXPY().

Referenced by default_axpy(), and default_axpy< float >().

template<>

void CTF_int::default_axpy< float >	(	int	n,
		float	alpha,
		float const *	X,
		int	incX,
		float *	Y,
		int	incY
	)

Definition at line 128 of file semiring.cxx.

References default_axpy< double >(), and CTF_BLAS::SAXPY().

Referenced by default_axpy().

template<>

void CTF_int::default_axpy< std::complex< double > >	(	int	n,
		std::complex< double >	alpha,
		std::complex< double > const *	X,
		int	incX,
		std::complex< double > *	Y,
		int	incY
	)

Definition at line 161 of file semiring.cxx.

References CTF_BLAS::ZAXPY().

template<>

void CTF_int::default_axpy< std::complex< float > >	(	int	n,
		std::complex< float >	alpha,
		std::complex< float > const *	X,
		int	incX,
		std::complex< float > *	Y,
		int	incY
	)

Definition at line 150 of file semiring.cxx.

References CTF_BLAS::CAXPY().

template<typename dtype >

void CTF_int::default_coomm	(	int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		int const *	rows_A,
		int const *	cols_A,
		int	nnz_A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 299 of file semiring.h.

References default_coomm< double >(), and default_coomm< float >().

Referenced by default_gemm_batch< std::complex< double > >().

template<>

void CTF_int::default_coomm< double >	(	int	m,
		int	n,
		int	k,
		double	alpha,
		double const *	A,
		int const *	rows_A,
		int const *	cols_A,
		int	nnz_A,
		double const *	B,
		double	beta,
		double *	C
	)

Definition at line 233 of file semiring.cxx.

References DEF_COOMM_KERNEL, and CTF_BLAS::MKL_DCOOMM().

Referenced by default_coomm(), and default_coomm< float >().

template<>

void CTF_int::default_coomm< float >	(	int	m,
		int	n,
		int	k,
		float	alpha,
		float const *	A,
		int const *	rows_A,
		int const *	cols_A,
		int	nnz_A,
		float const *	B,
		float	beta,
		float *	C
	)

Definition at line 208 of file semiring.cxx.

References DEF_COOMM_KERNEL, default_coomm< double >(), and CTF_BLAS::MKL_SCOOMM().

Referenced by default_coomm().

template<>

void CTF_int::default_coomm< std::complex< double > >	(	int	m,
		int	n,
		int	k,
		std::complex< double >	alpha,
		std::complex< double > const *	A,
		int const *	rows_A,
		int const *	cols_A,
		int	nnz_A,
		std::complex< double > const *	B,
		std::complex< double >	beta,
		std::complex< double > *	C
	)

Definition at line 286 of file semiring.cxx.

References DEF_COOMM_KERNEL, CTF_BLAS::MKL_ZCOOMM(), and muladd_csrmm().

template<>

void CTF_int::default_coomm< std::complex< float > >	(	int	m,
		int	n,
		int	k,
		std::complex< float >	alpha,
		std::complex< float > const *	A,
		int const *	rows_A,
		int const *	cols_A,
		int	nnz_A,
		std::complex< float > const *	B,
		std::complex< float >	beta,
		std::complex< float > *	C
	)

Definition at line 261 of file semiring.cxx.

References DEF_COOMM_KERNEL, and CTF_BLAS::MKL_CCOOMM().

template<typename dtype >

void CTF_int::default_fxpy	(	int	n,
		dtype const *	X,
		dtype *	Y
	)

Definition at line 19 of file monoid.h.

template<typename dtype >

void CTF_int::default_gemm	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 71 of file semiring.h.

template<>

void CTF_int::default_gemm< double >	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		double	alpha,
		double const *	A,
		double const *	B,
		double	beta,
		double *	C
	)

inline

Definition at line 166 of file semiring.h.

Referenced by default_gemm< float >(), and CTF::Semiring< dtype, is_ord >::is_offloadable().

template<>

void CTF_int::default_gemm< float >	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		float	alpha,
		float const *	A,
		float const *	B,
		float	beta,
		float *	C
	)

inline

Definition at line 151 of file semiring.h.

References default_gemm< double >().

Referenced by get_grp_ptrs(), and CTF::Semiring< dtype, is_ord >::is_offloadable().

template<>

void CTF_int::default_gemm< std::complex< double > >	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		std::complex< double >	alpha,
		std::complex< double > const *	A,
		std::complex< double > const *	B,
		std::complex< double >	beta,
		std::complex< double > *	C
	)

inline

Definition at line 196 of file semiring.h.

References default_gemm_batch().

template<>

void CTF_int::default_gemm< std::complex< float > >	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		std::complex< float >	alpha,
		std::complex< float > const *	A,
		std::complex< float > const *	B,
		std::complex< float >	beta,
		std::complex< float > *	C
	)

inline

Definition at line 181 of file semiring.h.

template<typename dtype >

void CTF_int::default_gemm_batch	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 211 of file semiring.h.

References default_gemm_batch< float >().

Referenced by default_gemm< std::complex< double > >().

template<>

void CTF_int::default_gemm_batch< double >	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		double	alpha,
		double const *	A,
		double const *	B,
		double	beta,
		double *	C
	)

inline

Definition at line 251 of file semiring.h.

Referenced by default_gemm_batch< float >().

template<>

void CTF_int::default_gemm_batch< float >	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		float	alpha,
		float const *	A,
		float const *	B,
		float	beta,
		float *	C
	)

inline

Definition at line 235 of file semiring.h.

References default_gemm_batch< double >().

Referenced by default_gemm_batch().

template<>

void CTF_int::default_gemm_batch< std::complex< double > >	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		std::complex< double >	alpha,
		std::complex< double > const *	A,
		std::complex< double > const *	B,
		std::complex< double >	beta,
		std::complex< double > *	C
	)

inline

Definition at line 283 of file semiring.h.

References default_coomm().

template<>

void CTF_int::default_gemm_batch< std::complex< float > >	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		std::complex< float >	alpha,
		std::complex< float > const *	A,
		std::complex< float > const *	B,
		std::complex< float >	beta,
		std::complex< float > *	C
	)

inline

Definition at line 267 of file semiring.h.

template<typename dtype , bool is_ord>

std::enable_if<is_ord, dtype>::type CTF_int::default_max	(	dtype	a,
		dtype	b
	)

inline

Definition at line 182 of file set.h.

References ctf.core::a.

template<typename dtype , bool is_ord>

std::enable_if<!is_ord, dtype>::type CTF_int::default_max	(	dtype	a,
		dtype	b
	)

inline

Definition at line 188 of file set.h.

References ctf.core::a.

template<typename dtype , bool is_ord>

std::enable_if<is_ord, dtype>::type CTF_int::default_max_lim ( )

inline

Definition at line 152 of file set.h.

template<typename dtype , bool is_ord>

std::enable_if<!is_ord, dtype>::type CTF_int::default_max_lim ( )

inline

Definition at line 158 of file set.h.

References ctf.core::a, and ctf.core::dtype.

template<typename dtype , bool is_ord>

std::enable_if<is_ord, dtype>::type CTF_int::default_min	(	dtype	a,
		dtype	b
	)

inline

Definition at line 138 of file set.h.

References ctf.core::a.

template<typename dtype , bool is_ord>

std::enable_if<!is_ord, dtype>::type CTF_int::default_min	(	dtype	a,
		dtype	b
	)

inline

Definition at line 144 of file set.h.

References ctf.core::a.

template<typename dtype , bool is_ord>

std::enable_if<is_ord, dtype>::type CTF_int::default_min_lim ( )

inline

Definition at line 167 of file set.h.

template<typename dtype , bool is_ord>

std::enable_if<!is_ord, dtype>::type CTF_int::default_min_lim ( )

inline

Definition at line 173 of file set.h.

References ctf.core::a, and ctf.core::dtype.

template<typename dtype >

dtype CTF_int::default_mul	(	dtype	a,
		dtype	b
	)

Definition at line 14 of file semiring.h.

References ctf.core::b.

template<typename dtype , void(*)(int, dtype const *, dtype *) fxpy>

void CTF_int::default_mxpy	(	void *	X,
		void *	Y,
		int *	n,
		MPI_Datatype *	d
	)

Definition at line 11 of file monoid.h.

References ctf.core::dtype.

Referenced by get_default_maddop().

template<typename dtype >

void CTF_int::default_scal	(	int	n,
		dtype	alpha,
		dtype *	X,
		int	incX
	)

Definition at line 47 of file semiring.h.

References default_scal< double >(), and default_scal< float >().

template<>

void CTF_int::default_scal< double >	(	int	n,
		double	alpha,
		double *	X,
		int	incX
	)

Definition at line 176 of file semiring.cxx.

References CTF_BLAS::DSCAL().

Referenced by default_scal().

template<>

void CTF_int::default_scal< float >	(	int	n,
		float	alpha,
		float *	X,
		int	incX
	)

Definition at line 171 of file semiring.cxx.

References CTF_BLAS::SSCAL().

Referenced by default_scal().

template<>

void CTF_int::default_scal< std::complex< double > >	(	int	n,
		std::complex< double >	alpha,
		std::complex< double > *	X,
		int	incX
	)

Definition at line 188 of file semiring.cxx.

References CTF_BLAS::ZSCAL().

template<>

void CTF_int::default_scal< std::complex< float > >	(	int	n,
		std::complex< float >	alpha,
		std::complex< float > *	X,
		int	incX
	)

Definition at line 182 of file semiring.cxx.

References CTF_BLAS::CSCAL().

void CTF_int::depad_tsr	(	int	order,
		int64_t	num_pair,
		int const *	edge_len,
		int const *	sym,
		int const *	padding,
		int const *	prepadding,
		char const *	pairsb,
		char *	new_pairsb,
		int64_t *	new_num_pair,
		algstrct const *	sr
	)

retrieves the unpadded pairs

Parameters

[in]	order	tensor dimension
[in]	num_pair	number of pairs
[in]	edge_len	tensor edge lengths
[in]	sym	symmetry types of tensor
[in]	padding	padding of tensor (included in edge_len)
[in]	prepadding	padding at start of tensor (included in edge_len)
[in]	pairsb	padded array of pairs
[out]	new_pairsb	unpadded pairs
[out]	new_num_pair	number of unpadded pairs
[in]	sr	algstrct defines sizeo of each pair

Definition at line 51 of file pad.cxx.

References alloc(), alloc_ptr(), AS, cdealloc(), CTF_int::ConstPairIterator::k(), SH, SY, TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write().

Referenced by read_loc_pairs(), CTF_int::tensor::slice(), and CTF_int::tensor::sparsify().

void CTF_int::depermute_keys	(	int	order,
		int	num_pair,
		int const *	edge_len,
		int const *	new_edge_len,
		int const	permutation,
		char *	pairs,
		algstrct const *	sr
	)

depermutes keys (apply P^T)

Parameters

[in]	order	tensor dimension
[in]	num_pair	number of pairs
[in]	edge_len	old nonpadded tensor edge lengths
[in]	new_edge_len	new nonpadded tensor edge lengths
[in]	permutation	permutation to apply to keys of each pair
[in,out]	pairs	the keys and values as pairs
[in]	sr	algstrct defining data type of array

Definition at line 99 of file sparse_rw.cxx.

References alloc(), ASSERT, cdealloc(), CTF_int::PairIterator::k(), MIN, CTF_int::algstrct::pair_size(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write_key().

Referenced by CTF_int::tensor::permute().

void CTF_int::depin	(	algstrct const *	sr,
		int	order,
		int const *	lens,
		int const *	divisor,
		int	nvirt,
		int const *	virt_dim,
		int const *	phys_rank,
		char *	X,
		int64_t &	new_nnz_B,
		int64_t *	nnz_blk,
		char *&	new_B,
		bool	check_padding
	)

depins keys of n pairs

Definition at line 883 of file algstrct.cxx.

References alloc(), alloc_ptr(), cdealloc(), CTF_int::PairIterator::k(), CTF_int::algstrct::pair_alloc(), CTF_int::algstrct::pair_size(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::spctr_pin_keys::run(), CTF_int::tspsum_pin_keys::run(), and CTF_int::algstrct::~algstrct().

void CTF_int::desymmetrize	(	tensor *	sym_tsr,
		tensor *	nonsym_tsr,
		bool	is_C
	)

unfolds the data of a tensor

Parameters

[in]	sym_tsr	starting symmetric tensor (where data starts)
[in]	nonsym_tsr	new tensor with a potentially unfolded symmetry
[in]	is_C	whether the tensor is an output of the operation

Definition at line 12 of file symmetrization.cxx.

References CTF_int::algstrct::addinv(), alloc(), CTF_int::algstrct::alloc(), alloc_ptr(), AS, CTF_int::tensor::calc_nvirt(), CTF_int::algstrct::cast_double(), cdealloc(), CTF_int::tensor::clear_mapping(), copy_mapping(), CTF_int::tensor::data, CTF_int::tensor::edge_map, CTF_int::accumulatable::el_size, CTF_int::scaling::execute(), CTF_int::summation::execute(), CTF_int::tensor::has_home, CTF_int::tensor::home_buffer, CTF_int::tensor::home_size, CTF_int::tensor::is_data_aliased, CTF_int::tensor::is_home, CTF_int::tensor::is_mapped, CTF_int::tensor::is_sparse, CTF_int::algstrct::mulid(), CTF_int::tensor::name, CTF_int::tensor::nnz_blk, NS, CTF_int::tensor::order, CTF_int::tensor::profile, CTF::World::rank, scal_diag(), CTF_int::tensor::set_new_nnz_glb(), CTF_int::tensor::set_padding(), CTF_int::tensor::set_zero(), SH, CTF_int::tensor::size, CTF_int::tensor::sr, CTF::Timer::start(), CTF::Timer::stop(), CTF_int::summation::sum_tensors(), SY, CTF_int::tensor::sym, TAU_FSTART, TAU_FSTOP, CTF_int::tensor::topo, VPRINTF, CTF_int::tensor::wrld, and CTF_int::tensor::zero_out_padding().

Referenced by CTF_int::summation::estimate_time(), and get_len_ordering().

std::vector<char> CTF_int::det_uniq_inds	(	std::vector< Term * > const	operands,
		std::vector< char > const	out_inds
	)

Definition at line 422 of file term.cxx.

Referenced by contract_down_terms(), CTF_int::Contract_Term::estimate_time(), CTF_int::Contract_Term::execute(), CTF_int::Sum_Term::get_uniq_inds(), and CTF_int::Contract_Term::get_uniq_inds().

double CTF_int::dgtog_est_time	(	int64_t	tot_sz,
		int	np
	)

estimates execution time, given this processor sends a receives tot_sz across np procs

Parameters

[in]	tot_sz	amount of data sent/recved
[in]	np	number of procs involved

Definition at line 11 of file dgtog_redist.cxx.

References dgtog_res_mdl.

Referenced by CTF_int::tensor::est_redist_time().

void CTF_int::dgtog_reshuffle	(	int const *	sym,
		int const *	edge_len,
		distribution const &	old_dist,
		distribution const &	new_dist,
		char **	ptr_tsr_data,
		char **	ptr_tsr_new_data,
		algstrct const *	sr,
		CommData	ord_glb_comm
	)

Definition at line 74 of file dgtog_redist.cxx.

Referenced by dgtog_reshuffle().

void CTF_int::dnA_spB_seq_sum	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		char const *	beta,
		char const *	B,
		int64_t	size_B,
		char *&	new_B,
		int64_t &	new_size_B,
		algstrct const *	sr_B,
		univar_function const *	func
	)

performs summation between two sparse tensors assumes B contain key value pairs sorted by key, with index permutation preapplied and with no repeated indices

Parameters

[in]	alpha	scaling factor of A
[in]	A	data of right operand
[in]	sr_A	algebraic structure of right operand
[in]	order_A	order of tensor A
[in]	edge_len_A	dimensions of tensor A
[in]	sym_A	symmetry relations of tensor A
[in]	beta	scaling factor of left operand
[in]	B	data of left operand
[in]	size_B	number of nonzero entries in left operand
[in,out]	new_B	new data of output
[in,out]	new_size_B	number of nonzero entries in output
[in]	sr_B	algebraic structure of left operand and output
[in]	func	function (or NULL) to apply to right operand

Definition at line 149 of file spr_seq_sum.cxx.

Referenced by CTF_int::seq_tsr_spsum::run().

void CTF_int::dump_all_models ( std::string path )

Definition at line 50 of file model.cxx.

References get_all_models().

Referenced by CTF_int::Model::dump_data(), and train_all().

double CTF_int::est_time_transp	(	int	order,
		int const *	new_order,
		int const *	edge_len,
		int	dir,
		algstrct const *	sr
	)

estimates time needed to transposes a non-symmetric (folded) tensor based on performance models

Parameters

[in]	order	dimension of tensor
[in]	new_order	new ordering of dimensions
[in]	edge_len	original edge lengths
[in]	dir	which way are we going?
[in]	sr	algstrct defining element size

Returns: estimated time in seconds

Definition at line 757 of file nosym_transp.cxx.

References long_contig_transp_mdl, non_contig_transp_mdl, and shrt_contig_transp_mdl.

Referenced by CTF_int::tensor::est_time_unfold(), CTF_int::summation::estimate_time(), and get_len_ordering().

double CTF_int::estimate_download_time ( int64_t size )

estimate time it takes to upload

double CTF_int::estimate_upload_time ( int64_t size )

estimate time it takes to download

void CTF_int::extract_free_comms	(	topology const *	topo,
		int	order_A,
		mapping const *	edge_map_A,
		int	order_B,
		mapping const *	edge_map_B,
		int &	num_sub_phys_dims,
		CommData **	psub_phys_comm,
		int **	pcomm_idx
	)

extracts the set of physical dimensions still available for mapping

Parameters

[in]	topo	topology
[in]	order_A	dimension of A
[in]	edge_map_A	mapping of A
[in]	order_B	dimension of B
[in]	edge_map_B	mapping of B
[out]	num_sub_phys_dims	number of free torus dimensions
[out]	psub_phys_comm	the torus dimensions
[out]	pcomm_idx	index of the free torus dimensions in the origin topology

Definition at line 628 of file topology.cxx.

References alloc_ptr(), CTF_int::mapping::cdt, CTF_int::mapping::child, CTF_int::topology::dim_comm, CTF_int::mapping::has_child, CTF_int::topology::order, PHYSICAL_MAP, and CTF_int::mapping::type.

Referenced by get_len_ordering().

int64_t CTF_int::fact ( int64_t n )

Definition at line 277 of file util.cxx.

Referenced by align_symmetric_indices(), chchoose(), choose(), coalesce_bwd(), and overcounting_factor().

void CTF_int::factorize	(	int	n,
		int *	nfactor,
		int **	factor
	)

computes the size of a tensor in packed symmetric layout

Parameters

[in]	n	a positive number
[out]	nfactor	number of factors in n
[out]	factor	array of length nfactor, corresponding to factorization of n

Definition at line 170 of file util.cxx.

References alloc().

Referenced by fft(), get_generic_topovec(), and get_phys_topo().

int CTF_int::find_topology	(	topology const *	topo,
		std::vector< topology * > &	topovec
	)

searches for an equivalent topology in avector of topologies

Parameters

[in]	topo	topology to match
[in]	topovec	vector of existing parameters

Returns: -1 if not found, otherwise index of first found topology

Definition at line 571 of file topology.cxx.

References CTF_int::topology::lens, and CTF_int::topology::order.

Referenced by peel_perm_torus(), peel_torus(), and CTF_int::tensor::set_distribution().

void CTF_int::flops_add ( int64_t n )

Definition at line 173 of file common.cxx.

int CTF_int::free_cond ( void * ptr )

int CTF_int::gcd	(	int	a,
		int	b
	)

inline

Definition at line 334 of file util.h.

References ctf.core::a.

Referenced by lcm().

template<typename dtype >

void CTF_int::gemm	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 82 of file semiring.cxx.

Referenced by CTF::Bivar_Kernel< dtype_A, dtype_B, dtype_C, f, g >::cgemm(), CTF::Semiring< dtype, is_ord >::gemm_batch(), get_grp_ptrs(), CTF_int::algstrct::has_mul(), and sym_seq_ctr_inr().

template<typename dtype >

void CTF_int::gemm_batch	(	char	taA,
		char	taB,
		int	l,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 15 of file semiring.cxx.

References ctf.core::dtype, and get_grp_ptrs().

Referenced by get_grp_ptrs(), and CTF_int::algstrct::has_mul().

std::vector<Model*>& CTF_int::get_all_models ( )

Definition at line 10 of file model.cxx.

Referenced by dump_all_models(), CTF_int::LinModel< nparam >::LinModel(), load_all_models(), print_all_models(), update_all_models(), and write_all_models().

std::vector< topology* > CTF_int::get_all_topos	(	CommData	cdt,
		int	n_uf,
		int const *	uniq_fact,
		int const *	mults,
		int	n_prepend,
		int const *	prelens
	)

computes all unique factorizations into non-primes each yielding a topology, prepending additional factors as specified

Parameters

[in]	cdt	global communicator
[in]	n_uf	number of unique prime factors
[in]	uniq_fact	list of prime factors
[in]	n_prepend	number of factors to prepend
[in]	mults	?
[in]	prelens	factors to prepend

Returns: lens vector of factorizations

Definition at line 410 of file topology.cxx.

References ASSERT, and CTF_int::topology::topology().

Referenced by get_generic_topovec().

int CTF_int::get_best_topo	(	int64_t	nvirt,
		int	topo,
		CommData	global_comm,
		int64_t	bcomm_vol = `0`,
		int64_t	bmemuse = `0`
	)

get the best topologoes (least nvirt) over all procs

Parameters

[in]	nvirt	best virtualization achieved by this proc
[in]	topo	topology index corresponding to best virtualization
[in]	global_comm	is the global communicator
[in]	bcomm_vol	best comm volume computed
[in]	bmemuse	best memory usage computed return virtualization factor

Definition at line 591 of file topology.cxx.

References ASSERT, and CTF_int::CommData::cm.

Referenced by CTF_int::scaling::execute(), CTF_int::summation::is_equal(), and CTF_int::tensor::set_zero().

void CTF_int::get_choice	(	int64_t	n,
		int64_t	k,
		int64_t	ch,
		int *	chs
	)

Definition at line 289 of file util.cxx.

References calc_idx_arr(), NS, and SH.

Referenced by coalesce_bwd(), and get_len_ordering().

int64_t CTF_int::get_coo_size	(	int64_t	nnz,
		int	val_size
	)

Definition at line 7 of file coo.cxx.

Referenced by CTF_int::COO_Matrix::COO_Matrix(), CTF_int::COO_Matrix::size(), and CTF_int::tensor::spmatricize().

int64_t CTF_int::get_csr_size	(	int64_t	nnz,
		int	nrow,
		int	val_size
	)

computes the size of a serialized CSR matrix

Parameters

[in]	nnz	number of nonzeros in matrix
[in]	nrow	number of rows in matrix
[in]	val_size	size of each matrix entry

Definition at line 8 of file csr.cxx.

References ALIGN.

Referenced by CTF_int::CSR_Matrix::CSR_Matrix(), CTF_int::CSR_Matrix::partition(), CTF_int::CSR_Matrix::size(), and CTF_int::tensor::spmatricize().

template<typename dtype >

constexpr bool CTF_int::get_default_is_ord ( )

Definition at line 234 of file set.h.

template<typename dtype >

MPI_Op CTF_int::get_default_maddop ( )

Definition at line 28 of file monoid.h.

References default_mxpy(), and ctf.core::dtype.

template<>

MPI_Op CTF_int::get_default_maddop< bool > ( )

inline

Definition at line 38 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< char > ( )

inline

Definition at line 37 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< double > ( )

inline

Definition at line 44 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< float > ( )

inline

Definition at line 43 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< int > ( )

inline

Definition at line 39 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< int64_t > ( )

inline

Definition at line 40 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< long double > ( )

inline

Definition at line 45 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< std::complex< double > > ( )

inline

Definition at line 47 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< std::complex< float > > ( )

inline

Definition at line 46 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< uint64_t > ( )

inline

Definition at line 42 of file monoid.h.

template<>

MPI_Op CTF_int::get_default_maddop< unsigned int > ( )

inline

Definition at line 41 of file monoid.h.

template<typename dtype >

MPI_Datatype CTF_int::get_default_mdtype ( bool & is_custom )

Definition at line 194 of file set.h.

References ctf.core::dtype, MPI_CTF_BOOL, MPI_CTF_DOUBLE_COMPLEX, and MPI_CTF_LONG_DOUBLE_COMPLEX.

template<>

MPI_Datatype CTF_int::get_default_mdtype< bool > ( bool & is_custom )

inline

Definition at line 207 of file set.h.

References MPI_CTF_BOOL.

template<>

MPI_Datatype CTF_int::get_default_mdtype< char > ( bool & is_custom )

inline

Definition at line 213 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< double > ( bool & is_custom )

inline

Definition at line 225 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< float > ( bool & is_custom )

inline

Definition at line 223 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< int > ( bool & is_custom )

inline

Definition at line 215 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< int64_t > ( bool & is_custom )

inline

Definition at line 217 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< long double > ( bool & is_custom )

inline

Definition at line 227 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< std::complex< double > > ( bool & is_custom )

inline

Definition at line 209 of file set.h.

References MPI_CTF_DOUBLE_COMPLEX.

template<>

MPI_Datatype CTF_int::get_default_mdtype< std::complex< float > > ( bool & is_custom )

inline

Definition at line 229 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< std::complex< long double > > ( bool & is_custom )

inline

Definition at line 211 of file set.h.

References MPI_CTF_LONG_DOUBLE_COMPLEX.

template<>

MPI_Datatype CTF_int::get_default_mdtype< uint64_t > ( bool & is_custom )

inline

Definition at line 221 of file set.h.

template<>

MPI_Datatype CTF_int::get_default_mdtype< unsigned int > ( bool & is_custom )

inline

Definition at line 219 of file set.h.

int CTF_int::get_distribution_size ( int order )

inline

Definition at line 13 of file distribution.h.

Referenced by CTF_int::distribution::distribution(), CTF_int::tensor::orient_subworld(), and CTF_int::distribution::serialize().

algstrct const * CTF_int::get_double_ring ( )

Definition at line 10 of file ring.cxx.

References double_ring.

Referenced by CTF_int::Term::operator double().

algstrct const * CTF_int::get_float_ring ( )

Definition at line 6 of file ring.cxx.

References float_ring.

Referenced by CTF_int::Term::operator float().

int64_t CTF_int::get_flops ( )

Definition at line 177 of file common.cxx.

References total_flop_count.

Referenced by CTF::Flop_counter::count(), CTF::Flop_counter::Flop_counter(), and CTF::Flop_counter::zero().

template<typename dtype >

const char* CTF_int::get_fmt ( )

return format string for templated type

Definition at line 6 of file graph_io_aux.cxx.

References IASSERT.

template<>

const char* CTF_int::get_fmt< double > ( )

inline

Definition at line 18 of file graph_io_aux.cxx.

template<>

const char* CTF_int::get_fmt< float > ( )

inline

Definition at line 13 of file graph_io_aux.cxx.

template<>

const char* CTF_int::get_fmt< int > ( )

inline

Definition at line 23 of file graph_io_aux.cxx.

template<>

const char* CTF_int::get_fmt< int64_t > ( )

inline

Definition at line 28 of file graph_io_aux.cxx.

Idx_Tensor* CTF_int::get_full_intm	(	Idx_Tensor &	A,
		Idx_Tensor &	B,
		std::vector< char >	out_inds,
		bool	create_dummy = `false`
	)

Definition at line 121 of file term.cxx.

References alloc(), cdealloc(), CTF::Idx_Tensor::idx_map, CTF::Idx_Tensor::is_intm, CTF_int::tensor::is_sparse, CTF_int::tensor::lens, NS, CTF_int::tensor::order, CTF::Idx_Tensor::parent, CTF_int::tensor::sr, CTF_int::tensor::sym, and CTF_int::tensor::wrld.

Referenced by contract_down_terms(), CTF_int::Sum_Term::estimate_time(), and CTF_int::Sum_Term::execute().

std::vector< topology * > CTF_int::get_generic_topovec ( CommData cdt )

computes all topology configurations given undelying physical topology information

Parameters

[in] cdt global communicator

Definition at line 449 of file topology.cxx.

References cdealloc(), DPRINTF, factorize(), get_all_topos(), CTF_int::CommData::np, CTF_int::CommData::rank, and CTF_int::topology::topology().

Referenced by CTF::World::~World().

int CTF_int::get_glb	(	int	i,
		int	s,
		int	t
	)

inline

Definition at line 9 of file dgtog_calc_cnt.cxx.

Referenced by calc_sy_pfx(), calc_sy_pfx< 1 >(), and precompute_offsets().

template<typename dtype >

dtype** CTF_int::get_grp_ptrs	(	int64_t	grp_sz,
		int64_t	ngrp,
		dtype const *	data
	)

Definition at line 110 of file semiring.h.

References alloc(), default_gemm< float >(), ctf.core::dtype, gemm(), and gemm_batch().

Referenced by gemm_batch().

template<typename dtype >

void CTF_int::get_imag	(	tensor *	A,
		tensor *	B
	)

function that get the imaginary part from complex numbers

Parameters

[in] A tensor, param[in] B tensor stores the imaginary part from tensor A

Returns: None

Definition at line 60 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::dtype, and CTF_int::tensor::order.

template void CTF_int::get_imag< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::get_imag< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

algstrct const * CTF_int::get_int64_t_ring ( )

Definition at line 18 of file ring.cxx.

References int64_t_ring.

Referenced by CTF_int::Term::operator int64_t().

algstrct const * CTF_int::get_int_ring ( )

Definition at line 14 of file ring.cxx.

References int_ring.

Referenced by CTF_int::Term::operator int().

void CTF_int::get_len_ordering	(	tensor const *	A,
		tensor const *	B,
		tensor const *	C,
		int const *	idx_A,
		int const *	idx_B,
		int const *	idx_C,
		int **	new_ordering_A,
		int **	new_ordering_B,
		int **	new_ordering_C
	)

find ordering of indices of tensor to reduce to DGEMM (A, B, and C may be permuted

Parameters

[in]	A	tensor 1
[in]	B	tensor 2
[in]	C	tensor 3
[in]	idx_A	indices of tensor 1
[in]	idx_B	indices of tensor 2
[in]	idx_C	indices of tensor 3
[out]	new_ordering_A	the new ordering for indices of A
[out]	new_ordering_B	the new ordering for indices of B
[out]	new_ordering_C	the new ordering for indices of C

Definition at line 451 of file contraction.cxx.

int CTF_int::get_loc	(	int	g,
		int	s,
		int	t
	)

inline

Definition at line 16 of file dgtog_calc_cnt.cxx.

Referenced by calc_cnt(), calc_cnt< 0 >(), calc_sy_pfx(), calc_sy_pfx< 1 >(), and precompute_offsets().

template<typename dtype >

MPI_Op CTF_int::get_maddop ( void(*)(int, dtype const *, dtype *) fxpy )

Definition at line 50 of file monoid.h.

bool CTF_int::get_mpi_dt	(	int64_t	count,
		int64_t	datum_size,
		MPI_Datatype &	dt
	)

gives a datatype for arbitrary datum_size, errors if exceeding 32-bits

Parameters

[in]	count	number of elements we want to communicate
[in]	datum_size	element size
[in]	dt	new datatype to pass to MPI routine

Returns: whether the datatype is custom and needs to be freed

Definition at line 587 of file common.cxx.

References ASSERT, and MPI_CTF_DOUBLE_COMPLEX.

Referenced by CTF_int::tspsum_replicate::run().

std::list<mem_loc>* CTF_int::get_mst ( )

Definition at line 163 of file memcontrol.cxx.

References mst.

int CTF_int::get_num_instances ( )

Definition at line 531 of file memcontrol.cxx.

References instance_counter.

Referenced by CTF::World::~World().

template<typename ptype >

void CTF_int::get_perm	(	int	perm_order,
		ptype	A,
		ptype	B,
		ptype	C,
		ptype &	tA,
		ptype &	tB,
		ptype &	tC
	)

Definition at line 117 of file contraction.cxx.

topology * CTF_int::get_phys_topo	(	CommData	glb_comm,
		TOPOLOGY	mach
	)

get dimension and torus lengths of specified topology

Parameters

[in]	glb_comm	communicator
[in]	mach	specified topology

Definition at line 94 of file topology.cxx.

References alloc(), cdealloc(), ctf.core::dim, factorize(), MIN, NO_TOPOLOGY, CTF_int::CommData::np, ctf.core::np(), CTF_int::topology::order, CTF_int::topology::topology(), TOPOLOGY_8D, TOPOLOGY_BGP, TOPOLOGY_BGQ, and TOPOLOGY_GENERIC.

Referenced by CTF::World::~World().

double CTF_int::get_rand48 ( )

returns new random number in [0,1)

Definition at line 27 of file common.cxx.

References allred_mdl, allred_mdl_cst, allred_mdl_cst_init, allred_mdl_init, alltoall_mdl, alltoall_mdl_init, alltoallv_mdl, alltoallv_mdl_init, bcast_mdl, bcast_mdl_init, red_mdl, red_mdl_cst, red_mdl_cst_init, red_mdl_init, and rng.

Referenced by CTF::fill_random_base(), CTF::fill_sp_random_base(), and subsample().

template<typename dtype >

void CTF_int::get_real	(	tensor *	A,
		tensor *	B
	)

function that get the real part from complex numbers

Parameters

[in] A tensor, param[in] B tensor stores the real part from tensor A

Returns: None

Definition at line 51 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::dtype, and CTF_int::tensor::order.

template void CTF_int::get_real< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::get_real< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

void CTF_int::get_sym_perms	(	summation const &	sum,
		std::vector< summation > &	perms,
		std::vector< int > &	signs
	)

finds all permutations of a summation that must be done for a broken symmetry

Parameters

[in]	sum	summation specification
[out]	perms	the permuted summation specifications
[out]	signs	sign of each summation

Definition at line 647 of file symmetrization.cxx.

References CTF_int::summation::A, add_sym_perm(), AS, CTF_int::summation::B, CTF_int::summation::idx_A, CTF_int::summation::idx_B, NS, CTF_int::tensor::order, sign(), and CTF_int::tensor::sym.

Referenced by CTF_int::summation::estimate_time(), and get_len_ordering().

void CTF_int::get_sym_perms	(	contraction const &	ctr,
		std::vector< contraction > &	perms,
		std::vector< int > &	signs
	)

finds all permutations of a contraction that must be done for a broken symmetry

Parameters

[in]	ctr	contraction specification
[out]	perms	the permuted contraction specifications
[out]	signs	sign of each contraction

Definition at line 691 of file symmetrization.cxx.

References CTF_int::contraction::A, add_sym_perm(), AS, CTF_int::contraction::B, CTF_int::contraction::C, CTF_int::contraction::idx_A, CTF_int::contraction::idx_B, CTF_int::contraction::idx_C, NS, CTF_int::tensor::order, sign(), and CTF_int::tensor::sym.

int64_t CTF_int::getTotalSystemMemory ( )

inline

Definition at line 155 of file util.h.

char * CTF_int::glb_cyclic_reshuffle	(	int const *	sym,
		distribution const &	old_dist,
		int const *	old_offsets,
		int const	old_permutation,
		distribution const &	new_dist,
		int const *	new_offsets,
		int const	new_permutation,
		char **	ptr_tsr_data,
		char **	ptr_tsr_cyclic_data,
		algstrct const *	sr,
		CommData	ord_glb_comm,
		bool	reuse_buffers,
		char const *	alpha,
		char const *	beta
	)

Goes from any set of phases to any new set of phases.

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	old_dist	starting data distrubtion
[in]	old_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	old_permutation	permutation array for each edge length (no perm if NULL)
[in]	new_dist	target data distrubtion
[in]	new_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	new_permutation	permutation array for each edge length (no perm if NULL)
[in]	ptr_tsr_data	starting data buffer
[out]	ptr_tsr_cyclic_data	target data buffer
[in]	sr	algstrct defining data
[in]	ord_glb_comm	communicator on which to redistribute
[in]	reuse_buffers	if 1: ptr_tsr_cyclic_data is allocated dynamically and ptr_tsr_data is overwritten with intermediate data if 0: ptr_tsr_cyclic_data is preallocated and can be scaled by beta, however, more memory is used for temp buffers
[in]	alpha	scaling tensor for new data
[in]	beta	scaling tensor for original data

Definition at line 879 of file glb_cyclic_reshuffle.cxx.

void CTF_int::glb_ord_pup	(	int const *	sym,
		distribution const &	old_dist,
		distribution const &	new_dist,
		int const *	len,
		int const *	old_phys_dim,
		int const *	old_phys_edge_len,
		int const *	old_virt_edge_len,
		int64_t	old_virt_nelem,
		int const *	old_offsets,
		int const	old_permutation,
		int	total_np,
		int const *	new_phys_dim,
		int const *	new_phys_edge_len,
		int const *	new_virt_edge_len,
		int64_t	new_virt_nelem,
		char *	old_data,
		char **	new_data,
		int	forward,
		int const	bucket_offset,
		char const *	alpha,
		char const *	beta,
		algstrct const *	sr
	)

void CTF_int::handler ( )

Definition at line 181 of file common.cxx.

References ctf.core::array(), and ctf.core::rank().

void CTF_int::host_pinned_alloc	(	void **	ptr,
		int64_t	size
	)

allocate a pinned host buffer

Parameters

[out]	ptr	pointer to define
[in]	size	amount of buffer space to allocate

Referenced by CTF_int::ctr_2d_general::run(), and CTF_int::spctr_2d_general::run().

void CTF_int::host_pinned_free ( void * ptr )

free a pinned host buffer

Parameters

[in] ptr pointer to free

Referenced by CTF_int::ctr_2d_general::run(), and CTF_int::spctr_2d_general::run().

bool CTF_int::hptt_is_applicable	(	int	order,
		int const *	new_order,
		int	elementSize
	)

Checks if the HPTT library is applicable.

Parameters

[in]	order	dimension of tensor
[in]	new_order	new ordering of dimensions
[in]	elementSize	element size

Definition at line 319 of file nosym_transp.cxx.

Referenced by nosym_transpose().

void CTF_int::inc_tot_mem_used ( int64_t a )

Definition at line 80 of file memcontrol.cxx.

References ctf.core::a, and ASSERT.

Referenced by CTF_int::tensor::deregister_size(), and CTF_int::tensor::register_size().

void CTF_int::init_rng ( int rank )

initialized random number generator

Parameters

[in] rank processor index

Definition at line 23 of file common.cxx.

References rng.

Referenced by ctf.random::all_seed(), ctf.random::seed(), and CTF::World::~World().

void CTF_int::inv_idx	(	int	order_A,
		int const *	idx_A,
		int *	order_tot,
		int **	idx_arr
	)

invert index map

Parameters

[in]	order_A	number of dimensions of A
[in]	idx_A	index map of A
[in]	order_B	number of dimensions of B
[in]	idx_B	index map of B
[out]	order_tot	number of total dimensions
[out]	idx_arr	2*order_tot index array
[in]	order_A	number of dimensions of A
[in]	idx_A	index map of A
[out]	order_tot	number of total dimensions
[out]	idx_arr	2*ndim_tot index array

Definition at line 19 of file sym_seq_scl.cxx.

References alloc().

void CTF_int::inv_idx	(	int	order_A,
		int const *	idx_A,
		int	order_B,
		int const *	idx_B,
		int	order_C,
		int const *	idx_C,
		int *	order_tot,
		int **	idx_arr
	)

invert index map

Parameters

[in]	order_A	number of dimensions of A
[in]	idx_A	index map of A
[in]	order_B	number of dimensions of B
[in]	idx_B	index map of B
[in]	order_C	number of dimensions of C
[in]	idx_C	index map of C
[out]	order_tot	number of total dimensions
[out]	idx_arr	3*order_tot index array

Definition at line 592 of file ctr_tsr.cxx.

References alloc(), CTF_int::seq_tsr_ctr::order_A, CTF_int::seq_tsr_ctr::order_B, and CTF_int::seq_tsr_ctr::order_C.

Referenced by add_sym_perm(), calc_fold_lnmk(), copy_mapping(), CTF_int::seq_tsr_spctr::est_fp(), CTF_int::seq_tsr_ctr::est_fp(), CTF_int::summation::estimate_time(), CTF_int::scaling::execute(), get_len_ordering(), CTF_int::summation::is_equal(), CTF_int::seq_tsr_scl::seq_tsr_scl(), spA_dnB_dnC_seq_ctr(), CTF_int::summation::sum_tensors(), sym_seq_ctr_cust(), sym_seq_ctr_inr(), sym_seq_ctr_ref(), sym_seq_scl_cust(), sym_seq_scl_ref(), sym_seq_sum_cust(), sym_seq_sum_inr(), sym_seq_sum_ref(), CTF_int::endomorphism::~endomorphism(), CTF_int::seq_tsr_ctr::~seq_tsr_ctr(), and CTF_int::seq_tsr_sum::~seq_tsr_sum().

void CTF_int::inv_idx	(	int	order_A,
		int const *	idx_A,
		int	order_B,
		int const *	idx_B,
		int *	order_tot,
		int **	idx_arr
	)

invert index map

Parameters

[in]	order_A	number of dimensions of A
[in]	idx_A	index map of A
[in]	order_B	number of dimensions of B
[in]	idx_B	index map of B
[out]	order_tot	number of total dimensions
[out]	idx_arr	2*order_tot index array

Definition at line 913 of file spsum_tsr.cxx.

References alloc().

int CTF_int::lcm	(	int	a,
		int	b
	)

inline

Definition at line 340 of file util.h.

References gcd().

Referenced by calc_drv_displs(), ctr_2d_gen_build(), dgtog_reshuffle(), get_len_ordering(), glb_ord_pup(), map_symtsr(), map_tensor(), and CTF_int::tensor::set_zero().

void CTF_int::lda_cpy	(	int	el_size,
		int	nrow,
		int	ncol,
		int	lda_A,
		int	lda_B,
		const char *	A,
		char *	B
	)

inline

Copies submatrix to submatrix (column-major)

Parameters

[in]	el_size	element size
[in]	nrow	number of rows
[in]	ncol	number of columns
[in]	lda_A	lda along rows for A
[in]	lda_B	lda along rows for B
[in]	A	matrix to read from
[in,out]	B	matrix to write to

Definition at line 355 of file util.h.

Referenced by CTF_int::LinModel< nparam >::update().

void CTF_int::load_all_models ( std::string file_name )

Definition at line 34 of file model.cxx.

References get_all_models().

Referenced by CTF_int::Model::dump_data(), and CTF::World::~World().

int CTF_int::map_self_indices	(	tensor const *	tsr,
		int const *	idx_map
	)

create virtual mapping for idx_maps that have repeating indices

Parameters

[in]	tsr	tensor handle
[in]	idx_map	mapping of tensor indices to contraction map

Definition at line 423 of file mapping.cxx.

References alloc_ptr(), cdealloc(), CTF_int::tensor::edge_map, CTF_int::mapping::has_child, map_symtsr(), NOT_MAPPED, CTF_int::mapping::np, CTF_int::tensor::order, SUCCESS, CTF_int::tensor::sym_table, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by CTF_int::scaling::execute(), and CTF_int::summation::is_equal().

int CTF_int::map_symtsr	(	int	tsr_order,
		int const *	tsr_sym_table,
		mapping *	tsr_edge_map
	)

adjust a mapping to maintan symmetry

Parameters

[in]	tsr_order	is the number of dimensions of the tensor
[in]	tsr_sym_table	the symmetry table of a tensor
[in,out]	tsr_edge_map	is the mapping

Returns: CTF::SUCCESS if mapping successful, CTF::NEGATIVE if not, CTF::ERROR if err'ed out

Definition at line 470 of file mapping.cxx.

References ASSERT, CTF_int::mapping::calc_phase(), CTF_int::mapping::child, CTF_int::mapping::has_child, lcm(), CTF_int::mapping::mapping(), MAX_PHASE, MAXLOOP, NEGATIVE, NOT_MAPPED, CTF_int::mapping::np, PHYSICAL_MAP, SUCCESS, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by get_len_ordering(), map_self_indices(), and map_tensor().

int CTF_int::map_tensor	(	int	num_phys_dims,
		int	tsr_order,
		int const *	tsr_edge_len,
		int const *	tsr_sym_table,
		int *	restricted,
		CommData *	phys_comm,
		int const *	comm_idx,
		int	fill,
		mapping *	tsr_edge_map
	)

map a tensor

Parameters

[in]	num_phys_dims	number of physical processor grid dimensions
[in]	tsr_order	number dims
[in]	tsr_edge_len	edge lengths of the tensor
[in]	tsr_sym_table	the symmetry table of a tensor
[in,out]	restricted	an array used to restricted the mapping of tensor dims
[in]	phys_comm	dimensional communicators
[in]	comm_idx	dimensional ordering
[in]	fill	if set does recursive mappings and uses all phys dims
[in,out]	tsr_edge_map	mapping of tensor

Returns: CTF_SUCCESS if mapping successful, CTF_NEGATIVE if not, CTF_ERROR if err'ed out

Definition at line 244 of file mapping.cxx.

References CTF_int::mapping::calc_phys_phase(), CTF_int::mapping::cdt, CTF_int::mapping::child, CTF_int::mapping::has_child, lcm(), map_symtsr(), CTF_int::mapping::mapping(), MAX_PHASE, NEGATIVE, NOT_MAPPED, CTF_int::mapping::np, CTF_int::CommData::np, PHYSICAL_MAP, SUCCESS, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by get_len_ordering(), CTF_int::summation::is_equal(), CTF_int::tensor::map_tensor_rem(), and CTF_int::tensor::set_zero().

void CTF_int::matrix_qr	(	tensor *	A,
		tensor *	Q,
		tensor *	R
	)

Definition at line 109 of file ctf_ext.cxx.

References CTF_int::accumulatable::el_size, CTF::Matrix< dtype >::qr(), and CTF_int::tensor::sr.

Referenced by ctf.core::MPI_Stop(), and ctf.core::qr().

void CTF_int::matrix_qr_cmplx	(	tensor *	A,
		tensor *	Q,
		tensor *	R
	)

Definition at line 141 of file ctf_ext.cxx.

References CTF_int::accumulatable::el_size, CTF::Matrix< dtype >::qr(), and CTF_int::tensor::sr.

Referenced by ctf.core::MPI_Stop(), and ctf.core::qr().

void CTF_int::matrix_svd	(	tensor *	A,
		tensor *	U,
		tensor *	S,
		tensor *	VT,
		int	rank
	)

Definition at line 175 of file ctf_ext.cxx.

References CTF_int::accumulatable::el_size, CTF_int::tensor::sr, and CTF::Matrix< dtype >::svd().

Referenced by ctf.core::MPI_Stop(), and ctf.core::svd().

void CTF_int::matrix_svd_cmplx	(	tensor *	A,
		tensor *	U,
		tensor *	S,
		tensor *	VT,
		int	rank
	)

Definition at line 213 of file ctf_ext.cxx.

References CTF_int::accumulatable::el_size, CTF_int::tensor::sr, and CTF::Matrix< dtype >::svd().

Referenced by ctf.core::MPI_Stop(), and ctf.core::svd().

void CTF_int::mem_create ( )

create instance of memory manager

Definition at line 187 of file memcontrol.cxx.

References max_threads.

Referenced by CTF::World::~World().

void CTF_int::mem_exit ( int rank )

exit instance of memory manager

Parameters

[in] rank processor index

Definition at line 207 of file memcontrol.cxx.

References max_threads, and mst_buffer_ptr.

Referenced by CTF::World::~World().

void CTF_int::morph_topo	(	topology const *	new_topo,
		topology const *	old_topo,
		int	order,
		mapping *	edge_map
	)

morphs a tensor topology into another

Parameters

[in]	new_topo	topology to change to
[in]	old_topo	topology we are changing from
[in]	order	number of tensor dimensions
[in,out]	edge_map	mapping whose topology mapping we are changing

Definition at line 700 of file topology.cxx.

References alloc_ptr(), ASSERT, cdealloc(), CTF_int::mapping::cdt, CTF_int::mapping::child, CTF_int::mapping::clear(), CTF_int::topology::dim_comm, CTF_int::mapping::has_child, CTF_int::topology::lda, CTF_int::mapping::np, CTF_int::CommData::np, ctf.core::np(), CTF_int::topology::order, PHYSICAL_MAP, CTF_int::mapping::type, and VIRTUAL_MAP.

Referenced by get_len_ordering().

void * CTF_int::mst_alloc ( int64_t const len )

mst_alloc allocates buffer on the specialized memory stack

Parameters

[in] len number of bytes

Definition at line 307 of file memcontrol.cxx.

References ASSERT, mst_alloc_ptr(), and SUCCESS.

Referenced by contract_mst(), and get_len_ordering().

int CTF_int::mst_alloc_ptr	(	int64_t const	len,
		void **const	ptr
	)

mst_alloc abstraction

Parameters

[in]	len	number of bytes
[in,out]	ptr	pointer to set to new allocation address

Definition at line 269 of file memcontrol.cxx.

References ALIGN_BYTES, alloc_ptr(), ASSERT, DPRINTF, CTF_int::mem_loc::len, MST_ALIGN_BYTES, mst_buffer_ptr, mst_buffer_used, CTF_int::mem_loc::ptr, and SUCCESS.

Referenced by CTF_int::CommData::all_to_allv(), bcast_step(), calc_cnt_displs(), cyclic_reshuffle(), glb_cyclic_reshuffle(), glb_ord_pup(), mst_alloc(), pad_cyclic_pup_virt_buff(), CTF_int::ctr_2d_general::run(), CTF_int::spctr_2d_general::run(), and CTF_int::spctr_pin_keys::run().

void CTF_int::mst_create ( int64_t size )

initializes stack buffer

Definition at line 170 of file memcontrol.cxx.

References ALIGN_BYTES, and ASSERT.

Referenced by CTF::World::~World().

int CTF_int::mst_free ( void * ptr )

frees buffer allocated on stack

Parameters

[in] ptr pointer to buffer on stack

Definition at line 234 of file memcontrol.cxx.

References ABORT, ASSERT, ERROR, ctf.core::it, and SUCCESS.

Referenced by cdealloc().

template<typename dtype >

void CTF_int::muladd_csrmm	(	int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	A,
		int const *	JA,
		int const *	IA,
		int	nnz_A,
		dtype const *	B,
		dtype	beta,
		dtype *	C
	)

Definition at line 332 of file semiring.cxx.

References ctf.core::dtype, and muladd_csrmultd().

Referenced by default_coomm< std::complex< double > >().

template<typename dtype >

void CTF_int::muladd_csrmultd	(	int	m,
		int	n,
		int	k,
		dtype const *	A,
		int const *	JA,
		int const *	IA,
		int	nnz_A,
		dtype const *	B,
		int const *	JB,
		int const *	IB,
		int	nnz_B,
		dtype *	C
	)

Definition at line 370 of file semiring.cxx.

References CTF::Semiring< dtype, is_ord >::default_csrmm().

Referenced by muladd_csrmm().

void CTF_int::nosym_transpose	(	tensor *	A,
		int	all_fdim_A,
		int const *	all_flen_A,
		int const *	new_order,
		int	dir
	)

void CTF_int::nosym_transpose	(	int	order,
		int const *	new_order,
		int const *	edge_len,
		char *	data,
		int	dir,
		algstrct const *	sr
	)

transposes a non-symmetric (folded) tensor

Parameters

[in]	order	dimension of tensor
[in]	new_order	new ordering of dimensions
[in]	edge_len	original edge lengths
[in,out]	data	data tp transpose
[in]	dir	which way are we going?
[in]	sr	algstrct defining element size

Definition at line 519 of file nosym_transp.cxx.

References alloc_ptr(), cdealloc(), CTF_int::accumulatable::el_size, MIN, nosym_transpose(), and TAU_FSTOP.

void CTF_int::nosym_transpose	(	int	order,
		int const *	new_order,
		int const *	edge_len,
		char const *	data,
		int	dir,
		int	max_ntd,
		char **	tswap_data,
		int64_t *	chunk_size,
		algstrct const *	sr
	)

transposes a non-symmetric (folded) tensor internal kernel

Parameters

[in]	order	dimension of tensor
[in]	new_order	new ordering of dimensions
[in]	edge_len	original edge lengths
[in]	data	data tp transpose
[in]	dir	which way are we going?
[in]	max_ntd	how many threads to use
[out]	tswap_data	tranposed data
[out]	chunk_size	chunk sizes of tranposed data
[in]	sr	algstrct defining element size

Definition at line 586 of file nosym_transp.cxx.

References alloc_ptr(), ASSERT, cdealloc(), CTF_int::algstrct::copy(), CTF_int::accumulatable::el_size, TAU_FSTART, and TAU_FSTOP.

void CTF_int::nosym_transpose_hptt	(	int	order,
		int const *	edge_len,
		int	dir,
		tensor *&	A
	)

High-performance implementation of nosym_transpose using HPTT.

Parameters

[in]	order	dimension of tensor
[in]	edge_len	original edge lengths
[in]	dir	which way are we going?
[in,out]	A	tensor to be transposed

void CTF_int::nosym_transpose_hptt	(	int	order,
		int const *	st_new_order,
		int const *	st_edge_len,
		int	dir,
		char const *	st_buffer,
		char *	new_buffer,
		algstrct const *	sr
	)

Definition at line 332 of file nosym_transp.cxx.

References ABORT, CTF_int::accumulatable::el_size, and MIN.

Referenced by nosym_transpose().

void CTF_int::offload_exit ( )

exit offloading, e.g. destroy cublas

Referenced by CTF::World::~World().

template<typename dtype >

void CTF_int::offload_gemm	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		offload_tsr &	A,
		int	lda_A,
		offload_tsr &	B,
		int	lda_B,
		dtype	beta,
		offload_tsr &	C,
		int	lda_C
	)

Referenced by CTF::Bivar_Kernel< dtype_A, dtype_B, dtype_C, f, g >::coffload_gemm(), CTF::Semiring< dtype, is_ord >::csrmultcsr(), and CTF_int::algstrct::has_mul().

template<typename dtype >

void CTF_int::offload_gemm	(	char	tA,
		char	tB,
		int	m,
		int	n,
		int	k,
		dtype	alpha,
		dtype const *	dev_A,
		int	lda_A,
		dtype const *	dev_B,
		int	lda_B,
		dtype	beta,
		dtype *	dev_C,
		int	lda_C
	)

void CTF_int::offload_init ( )

initialize offloading, e.g. create cublas

Referenced by CTF::World::~World().

template<int idim>

void CTF_int::ord_glb	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int	prev_idx = `0`,
		int64_t	glb_ord_offset = `0`,
		int64_t	blk_ord_offset = `0`
	)

Definition at line 533 of file glb_cyclic_reshuffle.cxx.

References NS, and CTF_int::distribution::virt_phase.

Referenced by ord_glb_omp().

template<>

void CTF_int::ord_glb< 0 >	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int	prev_idx,
		int64_t	glb_ord_offset,
		int64_t	blk_ord_offset
	)

inline

Definition at line 566 of file glb_cyclic_reshuffle.cxx.

References NS, and ord_glb< 7 >().

Referenced by ord_glb_omp< 0 >(), and order_globally().

template void CTF_int::ord_glb< 7 >	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int	prev_idx,
		int64_t	glb_ord_offset,
		int64_t	blk_ord_offset
	)

Referenced by ord_glb< 0 >().

template<int idim>

void CTF_int::ord_glb_omp	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int const *	idx_st,
		int const *	idx_end,
		int	prev_idx = `0`,
		int64_t	glb_ord_offset = `0`,
		int64_t	blk_ord_offset = `0`
	)

Definition at line 606 of file glb_cyclic_reshuffle.cxx.

References MIN, NS, ord_glb(), and CTF_int::distribution::virt_phase.

template<>

void CTF_int::ord_glb_omp< 0 >	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int const *	idx_st,
		int const *	idx_end,
		int	prev_idx,
		int64_t	glb_ord_offset,
		int64_t	blk_ord_offset
	)

Definition at line 655 of file glb_cyclic_reshuffle.cxx.

References ord_glb< 0 >(), and ord_glb_omp< 7 >().

template void CTF_int::ord_glb_omp< 7 >	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr,
		int const *	idx_st,
		int const *	idx_end,
		int	prev_idx,
		int64_t	glb_ord_offset,
		int64_t	blk_ord_offset
	)

Referenced by ord_glb_omp< 0 >().

void CTF_int::order_globally	(	int const *	sym,
		distribution const &	dist,
		int const *	virt_edge_len,
		int const *	virt_phase_lda,
		int64_t	vbs,
		bool	dir,
		char const *	tsr_data_in,
		char *	tsr_data_out,
		algstrct const *	sr
	)

reorder local buffer so that elements are in ordered according to where they are in the global tensor (interleave virtual blocks)

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	dist	distribution of data
[in]	virt_edge_len	dimensions of each block
[in]	virt_phase_lda	prefix sum of virtual blocks
[in]	vbs	size of virtual blocks
[in]	dir	if 1 then go to global layout, if 0 than from
[in]	tsr_data_in	starting data buffer
[out]	tsr_data_out	target data buffer
[in]	sr	algstrct defining data

Definition at line 689 of file glb_cyclic_reshuffle.cxx.

References alloc(), ASSERT, CASE_ORD_GLB, cdealloc(), CTF_int::algstrct::copy(), ctf.core::dim, CTF_int::accumulatable::el_size, MIN, NS, ord_glb< 0 >(), CTF_int::distribution::order, sy_calc_idx_arr(), TAU_FSTART, TAU_FSTOP, and CTF_int::distribution::virt_phase.

Referenced by glb_cyclic_reshuffle().

void CTF_int::order_perm	(	tensor const *	A,
		tensor const *	B,
		int *	idx_arr,
		int	off_A,
		int	off_B,
		int *	idx_A,
		int *	idx_B,
		int &	add_sign,
		int &	mod
	)

orders the summation indices of one tensor that don't break summation symmetries

Parameters

[in]	A
[in]	B
[in]	idx_arr	inverted summation index map
[in]	off_A	offset of A in inverted index map
[in]	off_B	offset of B in inverted index map
[in]	idx_A	index map of A
[in]	idx_B	index map of B
[in,out]	add_sign	sign of contraction
[in,out]	mod	1 if sum is permuted

Definition at line 440 of file symmetrization.cxx.

References AS, ASSERT, MAX, MIN, NS, CTF_int::tensor::order, and CTF_int::tensor::sym.

void CTF_int::order_perm	(	tensor const *	A,
		tensor const *	B,
		tensor const *	C,
		int *	idx_arr,
		int	off_A,
		int	off_B,
		int	off_C,
		int *	idx_A,
		int *	idx_B,
		int *	idx_C,
		int &	add_sign,
		int &	mod
	)

orders the contraction indices of one tensor that don't break contraction symmetries

Parameters

[in]	A
[in]	B
[in]	C
[in]	idx_arr	inverted contraction index map
[in]	off_A	offset of A in inverted index map
[in]	off_B	offset of B in inverted index map
[in]	off_C	offset of C in inverted index map
[in]	idx_A	index map of A
[in]	idx_B	index map of B
[in]	idx_C	index map of C
[in,out]	add_sign	sign of contraction
[in,out]	mod	1 if permutation done

Definition at line 492 of file symmetrization.cxx.

References AS, MAX, MIN, NS, CTF_int::tensor::order, and CTF_int::tensor::sym.

Referenced by add_sym_perm().

int64_t CTF_int::packed_size	(	int	order,
		const int *	len,
		const int *	sym
	)

computes the size of a tensor in packed symmetric (SY, SH, or AS) layout

Parameters

[in]	order	tensor dimension
[in]	len	tensor edge _elngths
[in]	sym	tensor symmetries

Returns: size of tensor in packed layout

Definition at line 38 of file util.cxx.

Referenced by calc_idx_arr(), CTF_int::tensor::get_tot_size(), glb_ord_pup(), pad_cyclic_pup_virt_buff(), CTF_int::tensor::read_dense_from_file(), and CTF_int::tensor::write_dense_to_file().

void CTF_int::pad_cyclic_pup_virt_buff	(	int const *	sym,
		distribution const &	old_dist,
		distribution const &	new_dist,
		int const *	len,
		int const *	old_phys_dim,
		int const *	old_phys_edge_len,
		int const *	old_virt_edge_len,
		int64_t	old_virt_nelem,
		int const *	old_offsets,
		int const	old_permutation,
		int	total_np,
		int const *	new_phys_dim,
		int const *	new_phys_edge_len,
		int const *	new_virt_edge_len,
		int64_t	new_virt_nelem,
		char *	old_data,
		char **	new_data,
		int	forward,
		int const	bucket_offset,
		char const *	alpha,
		char const *	beta,
		algstrct const *	sr
	)

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	old_dist	starting data distrubtion
[in]	new_dist	target data distrubtion
[in]	len	non-padded edge lengths of tensor
[in]	old_phys_dim	edge lengths of the old processor grid
[in]	old_phys_edge_len	the old tensor processor block lengths
[in]	old_virt_edge_len	the old tensor block lengths
[in]	old_virt_nelem	the old number of elements per block
[in]	old_offsets	old offsets of each tensor edge (corner 1 of slice)
[in]	old_permutation	permutation array for each edge length (no perm if NULL)
[in]	total_np	the total number of processors
[in]	new_phys_dim	edge lengths of the new processor grid
[in]	new_phys_edge_len	the new tensor processor block lengths
[in]	new_virt_edge_len	the new tensor block lengths
[in]	new_virt_nelem	the new number of elements per block
[in,out]	old_data	the previous set of values stored locally
[in,out]	new_data	buffers to fill with data to send to each process and virtual bucket
[in]	forward	is 0 on the receiving side and reverses the role of all the previous parameters
[in]	bucket_offset	offsets for target index for each dimension
[in]	alpha	scaling factor for received data
[in]	beta	scaling factor for previous data
[in]	sr	algstrct defining elements and ops

Definition at line 8 of file cyclic_reshuffle.cxx.

References ABORT, CTF_int::algstrct::acc(), CTF_int::algstrct::addid(), alloc_ptr(), ASSERT, calc_idx_arr(), cdealloc(), CTF_int::algstrct::copy(), ctf.core::dim, CTF_int::accumulatable::el_size, CTF_int::algstrct::isequal(), MAX, MIN, mst_alloc_ptr(), CTF_int::algstrct::mulid(), NS, CTF_int::distribution::order, packed_size(), CTF_int::distribution::perank, CTF_int::distribution::phase, CTF_int::distribution::phys_phase, ctf.core::rank(), SY, sy_packed_size(), TAU_FSTART, TAU_FSTOP, and CTF_int::distribution::virt_phase.

Referenced by cyclic_reshuffle().

void CTF_int::pad_key	(	int	order,
		int64_t	num_pair,
		int const *	edge_len,
		int const *	padding,
		PairIterator	pairs,
		algstrct const *	sr,
		int const *	offsets = `NULL`
	)

applies padding to keys

Parameters

[in]	order	tensor dimension
[in]	num_pair	number of pairs
[in]	edge_len	tensor edge lengths
[in]	padding	padding of tensor (included in edge_len)
[in,out]	pairs	set of pairs which to pad
[in]	sr	algstrct defines sizeo of each pair
[in]	offsets	(default NULL, none applied), offsets keys

Definition at line 6 of file pad.cxx.

References CTF_int::PairIterator::k(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write_key().

Referenced by read_loc_pairs(), CTF_int::tensor::slice(), CTF_int::tensor::sparsify(), and wr_pairs_layout().

void CTF_int::padded_reshuffle	(	int const *	sym,
		distribution const &	old_dist,
		distribution const &	new_dist,
		char *	tsr_data,
		char **	tsr_cyclic_data,
		algstrct const *	sr,
		CommData	ord_glb_comm
	)

Reshuffle elements using key-value pair read/write.

Parameters

[in]	sym	symmetry relations between tensor dimensions
[in]	old_dist	starting data distrubtion
[in]	new_dist	target data distrubtion
[in]	tsr_data	starting data buffer
[out]	tsr_cyclic_data	target data buffer
[in]	sr	algstrct defining data
[in]	ord_glb_comm	communicator on which to redistribute

Definition at line 8 of file redist.cxx.

Referenced by CTF_int::tensor::redistribute().

template<typename dtype >

void CTF_int::parse_sparse_tensor_data	(	char **	lvals,
		int	order,
		dtype const *	pmulid,
		int *	lens,
		int64_t	nvals,
		CTF::Pair< dtype > *	pairs,
		bool	with_vals
	)

parse string containing sparse tensor into data

Parameters

[in]	lvals	array of string, one per line/entry, formatted as i1, ..., i_order v or i1, ..., i_order if with_vals=false
[in]	order	num modes in tensor
[in]	pmulid	pointer to multiplicative identity, used only if with_vals=false
[in]	lens	dimensions of tensor
[in]	nvals	number of entries in lvals
[in]	pairs	array of tensor index/value pairs to fill
[in]	with_vals	whether values are included in file

Definition at line 45 of file graph_io_aux.cxx.

References CTF::Pair< dtype >::d, ctf.core::dtype, and CTF::Pair< dtype >::k.

std::vector< topology * > CTF_int::peel_perm_torus	(	topology *	phys_topology,
		CommData	cdt
	)

folds specified topology and all of its permutations into all configurations of lesser dimensionality

Parameters

[in]	phys_topology	topology to fold
[in]	cdt	global communicator

Definition at line 488 of file topology.cxx.

References find_topology(), CTF_int::topology::lens, CTF_int::topology::order, peel_torus(), and CTF_int::topology::topology().

Referenced by CTF::World::~World().

std::vector< topology * > CTF_int::peel_torus	(	topology const *	topo,
		CommData	glb_comm
	)

folds specified topology into all configurations of lesser dimensionality

Parameters

[in]	topo	topology to fold
[in]	glb_comm	global communicator

Definition at line 537 of file topology.cxx.

References alloc(), cdealloc(), find_topology(), CTF_int::topology::lens, CTF_int::topology::order, and CTF_int::topology::topology().

Referenced by peel_perm_torus().

void CTF_int::permute	(	int	order,
		int const *	perm,
		int *	arr
	)

permute an array

Parameters

order	number of elements
perm	permutation array
arr	array to permute

Definition at line 205 of file util.cxx.

References alloc_ptr(), and cdealloc().

Referenced by coalesce_bwd(), and ctf.core::MPI_Stop().

void CTF_int::permute_keys	(	int	order,
		int	num_pair,
		int const *	edge_len,
		int const *	new_edge_len,
		int const	permutation,
		char *	pairs,
		int64_t *	new_num_pair,
		algstrct const *	sr
	)

permutes keys

Parameters

[in]	order	tensor dimension
[in]	num_pair	number of pairs
[in]	edge_len	old nonpadded tensor edge lengths
[in]	new_edge_len	new nonpadded tensor edge lengths
[in]	permutation	permutation to apply to keys of each pair
[in,out]	pairs	the keys and values as pairs
[out]	new_num_pair	number of new pairs, since pairs are ignored if perm[i][j] == -1
[in]	sr	algstrct defining data type of array

Definition at line 10 of file sparse_rw.cxx.

References CTF_int::PairIterator::k(), MIN, CTF_int::algstrct::pair_alloc(), CTF_int::algstrct::pair_dealloc(), CTF_int::algstrct::pair_size(), CTF_int::algstrct::set_pair(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write().

Referenced by CTF_int::tensor::permute().

void CTF_int::permute_target	(	int	order,
		int const *	perm,
		int *	arr
	)

permutes a permutation array

Parameters

order	number of elements in perm
perm	permutation array
arr	permutation array to permute

Definition at line 222 of file util.cxx.

References alloc_ptr(), and cdealloc().

Referenced by coalesce_bwd(), CTF_int::summation::estimate_time(), and get_len_ordering().

template<typename dtype >

void CTF_int::pow_helper	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

pow function

Parameters

[in] A tensor, param[in] B tensor, param[in,out] C tensor, param[in] index of A, param[in] index of B, param[in] index of C

Returns: None

Definition at line 26 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::b, and ctf.core::dtype.

template void CTF_int::pow_helper< bool >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< double >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< float >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< int16_t >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< int32_t >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< int64_t >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< int8_t >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

Referenced by conv_type().

template void CTF_int::pow_helper< std::complex< double > >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

template void CTF_int::pow_helper< std::complex< float > >	(	tensor *	A,
		tensor *	B,
		tensor *	C,
		char const *	idx_A,
		char const *	idx_B,
		char const *	idx_C
	)

void CTF_int::precompute_offsets	(	distribution const &	old_dist,
		distribution const &	new_dist,
		int const *	sym,
		int const *	len,
		int const *	rep_phase,
		int const *	phys_edge_len,
		int const *	virt_edge_len,
		int const *	virt_dim,
		int const *	virt_lda,
		int64_t	virt_nelem,
		int **	pe_offset,
		int **	bucket_offset,
		int64_t **	data_offset,
		int **	ivmax_pre
	)

Definition at line 261 of file dgtog_calc_cnt.cxx.

References alloc_ptr(), ctf.core::dim, get_glb(), get_loc(), MAX, NS, CTF_int::distribution::order, CTF_int::distribution::pe_lda, CTF_int::distribution::perank, CTF_int::distribution::phase, CTF_int::distribution::phys_phase, SY, sy_packed_size(), TAU_FSTART, TAU_FSTOP, and CTF_int::distribution::virt_phase.

Referenced by dgtog_reshuffle().

void CTF_int::print_all_models ( )

Definition at line 23 of file model.cxx.

References get_all_models().

Referenced by CTF_int::Model::dump_data(), and CTF::print_timers().

int64_t CTF_int::proc_bytes_available ( )

gives total memory available on this MPI process

Definition at line 655 of file memcontrol.cxx.

References memcap, proc_bytes_total(), and proc_bytes_used().

Referenced by CTF_int::scaling::execute(), get_len_ordering(), CTF_int::tensor::set_zero(), CTF_int::tensor::write(), and CTF::World::~World().

int64_t CTF_int::proc_bytes_total ( )

gives total memory size per MPI process

Definition at line 604 of file memcontrol.cxx.

References MIN, and tot_mem_available.

Referenced by proc_bytes_available().

int64_t CTF_int::proc_bytes_used ( )

gives total memory used on this MPI process

Definition at line 538 of file memcontrol.cxx.

References tot_mem_used.

Referenced by main(), and proc_bytes_available().

template<typename dtype >

int64_t CTF_int::read_data_mpiio	(	CTF::World const *	dw,
		char const *	fpath,
		char ***	datastr
	)

read sparse tensor data from file using MPI-I/O, creating string with one entry per line (different entries on each process)

Parameters

[in]	dw	MPI world/comm
[in]	fpath	file name
[in]	datastr	array of strings to create and read from file

Definition at line 142 of file graph_io_aux.cxx.

References alloc(), cdealloc(), CTF::World::np, and CTF::World::rank.

void CTF_int::read_loc_pairs	(	int	order,
		int64_t	nval,
		int	num_virt,
		int const *	sym,
		int const *	edge_len,
		int const *	padding,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_phase,
		int *	phase_rank,
		int64_t *	nread,
		char const *	data,
		char **	pairs,
		algstrct const *	sr
	)

read tensor pairs local to processor

Parameters

[in]	order	tensor dimension
[in]	nval	number of local values
[in]	num_virt	new total virtualization factor
[in]	sym	symmetries of tensor
[in]	edge_len	tensor edge lengths
[in]	padding	padding of tensor
[in]	phase	blocking factor in each dimension
[in]	phys_phase	number of procs in each dimension
[in]	virt_phase	virtualization in each dimension
[in]	phase_rank	virtualized rank in total phase
[out]	nread	number of local pairs read
[in]	data	tensor data to read from
[out]	pairs	local pairs read
[in]	sr	algstrct context defining values

Definition at line 1235 of file sparse_rw.cxx.

References alloc_ptr(), assign_keys(), cdealloc(), depad_tsr(), pad_key(), CTF_int::algstrct::pair_alloc(), and CTF_int::algstrct::pair_dealloc().

Referenced by padded_reshuffle(), and CTF_int::tensor::read_local().

void CTF_int::readwrite	(	int	order,
		int64_t	size,
		char const *	alpha,
		char const *	beta,
		int	nvirt,
		int const *	edge_len,
		int const *	sym,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_dim,
		int *	phase_rank,
		char *	vdata,
		char *	pairs,
		char	rw,
		algstrct const *	sr
	)

read or write pairs from / to tensor

Parameters

[in]	order	tensor dimension
[in]	size	number of pairs
[in]	alpha	multiplier for new value
[in]	beta	multiplier for old value
[in]	nvirt	num local blocks
[in]	edge_len	tensor edge lengths
[in]	sym	symmetries of tensor
[in]	phase	total phase in each dimension
[in]	phys_phase	physical distribution phase
[in]	virt_dim	virtualization in each dimension
[in]	phase_rank	virtualized rank in total phase
[in,out]	vdata	data to read from or write to
[in,out]	pairs	pairs to read or write
[in]	rw	whether to read 'r' or write 'w'
[in]	sr	algstrct context defining values

Definition at line 685 of file sparse_rw.cxx.

References CTF_int::algstrct::add(), alloc_ptr(), ASSERT, cdealloc(), CTF_int::algstrct::copy(), CTF_int::PairIterator::d(), CTF_int::accumulatable::el_size, CTF_int::algstrct::mul(), NS, CTF_int::algstrct::pair_size(), CTF_int::PairIterator::read_val(), TAU_FSTART, TAU_FSTOP, and CTF_int::PairIterator::write_val().

Referenced by wr_pairs_layout().

template<int idim>

void CTF_int::redist_bucket	(	int const	bucket_offset,
		int64_t const	data_offset,
		int const	ivmax_pre,
		int	rep_phase0,
		int	virt_dim0,
		bool	data_to_buckets,
		char *__restrict__	data,
		char **__restrict__	buckets,
		int64_t *	counts,
		algstrct const *	sr,
		int64_t	data_off,
		int	bucket_off,
		int	prev_idx
	)

Definition at line 7 of file dgtog_bucket.h.

Referenced by dgtog_reshuffle().

template<>

void CTF_int::redist_bucket< 0 >	(	int const	bucket_offset,
		int64_t const	data_offset,
		int const	ivmax_pre,
		int	rep_phase0,
		int	virt_dim0,
		bool	data_to_buckets,
		char *__restrict__	data,
		char **__restrict__	buckets,
		int64_t *	counts,
		algstrct const *	sr,
		int64_t	data_off,
		int	bucket_off,
		int	prev_idx
	)

Definition at line 30 of file dgtog_bucket.h.

Referenced by isendrecv< 0 >().

void CTF_int::redist_bucket_r0	(	int const	bucket_offset,
		int64_t const	data_offset,
		int const	ivmax_pre,
		int	rep_phase0,
		int	rep_idx0,
		int	virt_dim0,
		bool	data_to_buckets,
		char *__restrict__	data,
		char **__restrict__	buckets,
		int64_t *	counts,
		algstrct const *	sr,
		int64_t	data_off,
		int	bucket_off,
		int	prev_idx
	)

Definition at line 90 of file dgtog_bucket.h.

References CTF_int::algstrct::copy(), and CTF_int::accumulatable::el_size.

Referenced by isendrecv< 0 >().

void CTF_int::reduce_step_post	(	int	edge_len,
		char *	C,
		bool	is_sparse_C,
		bool	move_C,
		algstrct const *	sr_C,
		int64_t	b_C,
		int64_t	s_C,
		char *	buf_C,
		CommData *	cdt_C,
		int64_t	ctr_sub_lda_C,
		int64_t	ctr_lda_C,
		int	nblk_C,
		int64_t *	size_blk_C,
		int &	new_nblk_C,
		int64_t *&	new_size_blk_C,
		int64_t *	offsets_C,
		int	ib,
		char const *&	rec_beta,
		char const *	beta,
		char *&	up_C,
		char *&	new_C,
		int	n_new_C_grps,
		int &	i_new_C_grp,
		char **	new_C_grps
	)

char* CTF_int::reduce_step_pre	(	int	edge_len,
		char *	C,
		bool	is_sparse_C,
		bool	move_C,
		algstrct const *	sr_C,
		int64_t	b_C,
		int64_t	s_C,
		char *	buf_C,
		CommData *	cdt_C,
		int64_t	ctr_sub_lda_C,
		int64_t	ctr_lda_C,
		int	nblk_C,
		int64_t const *	size_blk_C,
		int &	new_nblk_C,
		int64_t *&	new_size_blk_C,
		int64_t *	offsets_C,
		int	ib,
		char const *&	rec_beta
	)

Definition at line 244 of file spctr_2d_general.cxx.

References CTF_int::algstrct::addid(), alloc_ptr(), CTF_int::spctr_2d_general::ctr_sub_lda_C, CTF_int::spctr_2d_general::edge_len, and CTF_int::accumulatable::el_size.

Referenced by CTF_int::spctr_2d_general::run().

void CTF_int::scal_diag	(	int	order,
		int64_t	size,
		int	nvirt,
		int const *	edge_len,
		int const *	sym,
		int const *	padding,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_phase,
		int const *	cphase_rank,
		char *	vdata,
		algstrct const *	sr,
		int const *	sym_mask
	)

scales each element by 1/(number of entries equivalent to it after permutation of indices for which sym_mask is 1)

Parameters

[in]	order	tensor dimension
[in]	size	number of values
[in]	nvirt	total virtualization factor
[in]	edge_len	tensor edge lengths with padding
[in]	sym	symmetries of tensor
[in]	padding	how much of the edge lengths is padding
[in]	phase	phase of the tensor on virtualized processor grid
[in]	phys_phase	phase of the tensor on virtualized processor grid
[in]	virt_phase	virtual phase in each dimension
[in]	cphase_rank	physical phase rank
[in,out]	vdata	array of all local data
[in]	sr	algstrct defines sizeo of each pair
[in]	sym_mask	identifies which tensor indices are part of the symmetric group which diagonals we want to scale (i.e. sym_mask [1,1] does A["ii"]= (1./2.)*A["ii"])

Definition at line 567 of file pad.cxx.

References alloc_ptr(), ASSERT, calc_idx_arr(), CTF_int::algstrct::cast_double(), cdealloc(), ctf.core::dim, CTF_int::accumulatable::el_size, MIN, NS, CTF_int::algstrct::scal(), SY, TAU_FSTART, and TAU_FSTOP.

Referenced by desymmetrize(), CTF_int::tensor::scale_diagonals(), and symmetrize().

template<typename dtype >

void CTF_int::seq_coo_to_csr	(	int64_t	nz,
		int	nrow,
		dtype *	csr_vs,
		int *	csr_ja,
		int *	csr_ia,
		dtype const *	coo_vs,
		int const *	coo_rs,
		int const *	coo_cs
	)

Definition at line 17 of file set.h.

References ctf.core::a, ctf.core::b, ctf.core::dtype, and try_mkl_coo_to_csr().

template<typename dtype >

void CTF_int::seq_csr_to_coo	(	int64_t	nz,
		int	nrow,
		dtype const *	csr_vs,
		int const *	csr_ja,
		int const *	csr_ia,
		dtype *	coo_vs,
		int *	coo_rs,
		int *	coo_cs
	)

Definition at line 85 of file set.h.

References ctf.core::b, ctf.core::copy(), ctf.core::dtype, and try_mkl_csr_to_coo().

template<typename dtype >

char* CTF_int::serialize_sparse_tensor_data	(	int	order,
		int *	lens,
		int64_t	nvals,
		CTF::Pair< dtype > *	pairs,
		bool	with_vals,
		int64_t &	str_len
	)

serialize sparse tensor data to create string

Returns: lvals array of string, one per line/entry, formatted as i1, ..., i_order v or i1, ..., i_order if with_vals=false

Parameters

[in]	order	num modes in tensor
[in]	pmulid	pointer to multiplicative identity, used only if with_vals=false
[in]	lens	dimensions of tensor
[in]	nvals	number of entries in lvals
[in]	pairs	array of tensor index/value pairs to fill
[in]	with_vals	whether values are included in file
[out]	length	of string output

Definition at line 91 of file graph_io_aux.cxx.

References alloc(), and CTF::Pair< dtype >::k.

template<typename dtype >

void CTF_int::set_imag	(	tensor *	A,
		tensor *	B
	)

function that set the imaginary part from complex numbers

Parameters

[in] A tensor, param[in] B tensor stores the imaginary part from tensor A

Returns: None

Definition at line 79 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::b, ctf.core::dtype, and CTF_int::tensor::order.

template void CTF_int::set_imag< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::set_imag< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

void CTF_int::set_mem_size ( int64_t size )

sets what fraction of the memory capacity CTF can use

Definition at line 110 of file memcontrol.cxx.

Referenced by CTF::World::~World().

void CTF_int::set_memcap ( double cap )

sets what fraction of the memory capacity CTF can use

Parameters

[in] cap memory fraction

Definition at line 118 of file memcontrol.cxx.

Referenced by CTF::World::~World().

template<typename dtype >

void CTF_int::set_real	(	tensor *	A,
		tensor *	B
	)

function that set the real part from complex numbers

Parameters

[in] A tensor, param[in] B tensor stores the real part from tensor A

Returns: None

Definition at line 70 of file ctf_ext.cxx.

References ctf.core::a, ctf.core::b, ctf.core::dtype, and CTF_int::tensor::order.

template void CTF_int::set_real< double >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

template void CTF_int::set_real< float >	(	tensor *	A,
		tensor *	B
	)

Referenced by conv_type().

void CTF_int::socopy	(	int64_t	m,
		int64_t	n,
		int64_t	lda_a,
		int64_t	lda_b,
		int64_t const *	sizes_a,
		int64_t *&	sizes_b,
		int64_t *&	offsets_b
	)

Definition at line 240 of file util.cxx.

References alloc().

Referenced by bcast_step(), and coalesce_bwd().

void CTF_int::sp_read	(	algstrct const *	sr,
		int64_t	ntsr,
		ConstPairIterator	prs_tsr,
		char const *	alpha,
		int64_t	nread,
		PairIterator	prs_read,
		char const *	beta
	)

reads elements of a sparse set defining the tensor, into a sparse read set with potentially repeating keys

Parameters

[in]	sr	algstrct defining data type of array
[in]	ntsr	number of elements in sparse tensor
[in]	prs_tsr	pairs of the sparse tensor
[in]	alpha	scaling factor for data of the sparse tensor
[in]	nread	number of elements in the read set
[in,out]	prs_read	pairs of the read set
[in]	beta	scaling factor for data of the read set

Definition at line 1311 of file sparse_rw.cxx.

References ctf.core::a, CTF_int::algstrct::add(), CTF_int::algstrct::addid(), ctf.core::b, CTF_int::accumulatable::el_size, CTF_int::algstrct::mul(), CTF_int::ConstPairIterator::read_val(), CTF_int::PairIterator::read_val(), and CTF_int::PairIterator::write_val().

Referenced by wr_pairs_layout().

void CTF_int::sp_write	(	int	num_virt,
		algstrct const *	sr,
		int64_t *	vntsr,
		ConstPairIterator	vprs_tsr,
		char const *	beta,
		int64_t *	vnwrite,
		ConstPairIterator	vprs_write,
		char const *	alpha,
		int64_t *	vnnew,
		char *&	pprs_new
	)

writes pairs in a sparse write set to the sparse set of elements defining the tensor, resulting in a set of size between ntsr and ntsr+nwrite

Parameters

[in]	num_virt	num local blocks
[in]	sr	algstrct defining data type of array
[in]	vntsr	number of elements in sparse tensor
[in]	vprs_tsr	pairs of the sparse tensor
[in]	beta	scaling factor for data of the sparse tensor
[in]	vnwrite	number of elements in the write set
[in]	vprs_write	pairs of the write set
[in]	alpha	scaling factor for data of the write set
[out]	vnnew	number of elements in resulting set
[out]	pprs_new	char array containing the pairs of the resulting set

Definition at line 1362 of file sparse_rw.cxx.

References ctf.core::a, CTF_int::algstrct::add(), ctf.core::b, CTF_int::accumulatable::el_size, CTF_int::ConstPairIterator::k(), CTF_int::algstrct::mul(), CTF_int::algstrct::pair_alloc(), CTF_int::ConstPairIterator::read_val(), ctf.core::w, CTF_int::PairIterator::write(), and CTF_int::PairIterator::write_val().

Referenced by wr_pairs_layout().

template<int idim>

void CTF_int::spA_dnB_dnC_ctrloop	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int64_t const *	lda_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 12 of file sp_seq_ctr.cxx.

References ASSERT, NS, and CTF_int::ConstPairIterator::ptr.

Referenced by spA_dnB_dnC_seq_ctr().

template<>

void CTF_int::spA_dnB_dnC_ctrloop< 0 >	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int64_t const *	lda_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 153 of file sp_seq_ctr.cxx.

References ASSERT, CTF_FLOPS_ADD, NS, and spA_dnB_dnC_ctrloop< MAX_ORD >().

template<>

void CTF_int::spA_dnB_dnC_ctrloop< MAX_ORD >	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int64_t const *	lda_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Referenced by spA_dnB_dnC_ctrloop< 0 >().

void CTF_int::spA_dnB_dnC_seq_ctr	(	char const *	alpha,
		char const *	A,
		int64_t	size_A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		bivar_function const *	func
	)

Definition at line 358 of file sp_seq_ctr.cxx.

References CTF_int::algstrct::add(), CTF_int::algstrct::addid(), alloc(), ASSERT, cdealloc(), compute_syoffs(), CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, inv_idx(), CTF_int::algstrct::isequal(), MAX_ORD, CTF_int::algstrct::mul(), CTF_int::algstrct::mulid(), CTF_int::algstrct::scal(), CTF_int::algstrct::set(), spA_dnB_dnC_ctrloop(), SWITCH_ORD_CALL, sy_packed_size(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_spctr::run().

void CTF_int::spA_dnB_seq_sum	(	char const *	alpha,
		char const *	A,
		int64_t	size_A,
		algstrct const *	sr_A,
		char const *	beta,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		univar_function const *	func
	)

performs summation between two sparse tensors assumes A contains key value pairs sorted by key, with index permutation preapplied and with no repeated indices

Parameters

[in]	alpha	scaling factor of A
[in]	A	data of right operand
[in]	size_A	number of nonzero entries in right operand
[in]	sr_A	algebraic structure of right operand
[in]	beta	scaling factor of left operand
[in,out]	B	data of left operand
[in]	sr_B	algebraic structure of left operand and output
[in]	order_B	order of tensor B
[in]	edge_len_B	dimensions of tensor B
[in]	sym_B	symmetry relations of tensor B
[in]	func	function (or NULL) to apply to right operand

Definition at line 95 of file spr_seq_sum.cxx.

References CTF_int::univar_function::acc_f(), CTF_int::algstrct::add(), CTF_int::algstrct::addid(), ASSERT, CTF_int::ConstPairIterator::d(), CTF_int::accumulatable::el_size, CTF_int::algstrct::isequal(), MAX_ORD, CTF_int::algstrct::mul(), CTF_int::algstrct::mulid(), CTF_int::algstrct::scal(), CTF_int::algstrct::set(), spA_dnB_seq_sum_loop(), SWITCH_ORD_CALL, sy_packed_size(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_spsum::run().

template<int idim>

void CTF_int::spA_dnB_seq_sum_loop	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		char const *	beta,
		char *&	B,
		algstrct const *	sr_B,
		int	order_B,
		int64_t	idx_B,
		int const *	edge_len_B,
		int64_t const *	lda_B,
		int const *	sym_B,
		univar_function const *	func
	)

Definition at line 7 of file spr_seq_sum.cxx.

References NS.

Referenced by spA_dnB_seq_sum().

template<>

void CTF_int::spA_dnB_seq_sum_loop< 0 >	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		char const *	beta,
		char *&	B,
		algstrct const *	sr_B,
		int	order_B,
		int64_t	idx_B,
		int const *	edge_len_B,
		int64_t const *	lda_B,
		int const *	sym_B,
		univar_function const *	func
	)

Definition at line 33 of file spr_seq_sum.cxx.

References NS, and spA_dnB_seq_sum_loop< MAX_ORD >().

template void CTF_int::spA_dnB_seq_sum_loop< MAX_ORD >	(	char const *	alpha,
		ConstPairIterator &	A,
		int64_t &	size_A,
		algstrct const *	sr_A,
		char const *	beta,
		char *&	B,
		algstrct const *	sr_B,
		int	order_B,
		int64_t	idx_B,
		int const *	edge_len_B,
		int64_t const *	lda_B,
		int const *	sym_B,
		univar_function const *	func
	)

Referenced by spA_dnB_seq_sum_loop< 0 >().

void CTF_int::spA_spB_seq_sum	(	char const *	alpha,
		char const *	A,
		int64_t	size_A,
		algstrct const *	sr_A,
		char const *	beta,
		char *	B,
		int64_t	size_B,
		char *&	new_B,
		int64_t &	new_size_B,
		algstrct const *	sr_B,
		univar_function const *	func,
		int64_t	map_pfx
	)

performs summation between two sparse tensors assumes A and B contain key value pairs sorted by key, with index permutation preapplied and with no repeated indices

Parameters

[in]	alpha	scaling factor of A
[in]	A	data of right operand
[in]	size_A	number of nonzero entries in right operand
[in]	sr_A	algebraic structure of right operand
[in]	beta	scaling factor of left operand
[in]	B	data of left operand
[in]	size_B	number of nonzero entries in left operand
[in,out]	new_B	new data of output
[in,out]	new_size_B	number of nonzero entries in output
[in]	sr_B	algebraic structure of left operand and output
[in]	func	function (or NULL) to apply to right operand
[in]	map_pfx	how many times each element of A should be replicated

Definition at line 327 of file spr_seq_sum.cxx.

References spspsum().

Referenced by CTF_int::seq_tsr_spsum::run(), and spspsum().

void CTF_int::spcopy	(	int64_t	m,
		int64_t	n,
		int64_t	lda_a,
		int64_t	lda_b,
		int64_t const *	sizes_a,
		int64_t const *	offsets_a,
		char const *	a,
		int64_t const *	sizes_b,
		int64_t const *	offsets_b,
		char *	b
	)

Definition at line 260 of file util.cxx.

Referenced by bcast_step(), coalesce_bwd(), and CTF_int::spctr_2d_general::run().

int CTF_int::spctr_2d_gen_build	(	int	is_used,
		CommData	global_comm,
		int	i,
		int *	virt_dim,
		int &	cg_edge_len,
		int &	total_iter,
		tensor *	A,
		int	i_A,
		CommData *&	cg_cdt_A,
		int64_t &	cg_spctr_lda_A,
		int64_t &	cg_spctr_sub_lda_A,
		bool &	cg_move_A,
		int *	blk_len_A,
		int64_t &	blk_sz_A,
		int const *	virt_blk_len_A,
		int &	load_phase_A,
		tensor *	B,
		int	i_B,
		CommData *&	cg_cdt_B,
		int64_t &	cg_spctr_lda_B,
		int64_t &	cg_spctr_sub_lda_B,
		bool &	cg_move_B,
		int *	blk_len_B,
		int64_t &	blk_sz_B,
		int const *	virt_blk_len_B,
		int &	load_phase_B,
		tensor *	C,
		int	i_C,
		CommData *&	cg_cdt_C,
		int64_t &	cg_spctr_lda_C,
		int64_t &	cg_spctr_sub_lda_C,
		bool &	cg_move_C,
		int *	blk_len_C,
		int64_t &	blk_sz_C,
		int const *	virt_blk_len_C,
		int &	load_phase_C
	)

double CTF_int::spredist_est_time	(	int64_t	size,
		int	np
	)

Definition at line 24 of file untyped_tensor.cxx.

References spredist_mdl.

Referenced by CTF_int::tensor::est_redist_time().

void CTF_int::spsfy_tsr	(	int	order,
		int64_t	size,
		int	nvirt,
		int const *	edge_len,
		int const *	sym,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_dim,
		int *	phase_rank,
		char const *	vdata,
		char *&	vpairs,
		int64_t *	nnz_blk,
		algstrct const *	sr,
		int64_t const *	edge_lda,
		std::function< bool(char const *)>	f
	)

extracts all tensor values (in pair format) that pass a sparsifier function (including padded zeros if they pass the fliter)

Parameters

[in]	order	tensor dimension
[in]	size	number of values
[in]	nvirt	total virtualization factor
[in]	edge_len	tensor edge lengths (padded)
[in]	sym	symmetries of tensor
[in]	phase	total phase of the tensor on virtualized processor grid
[in]	phys_phase	physical phase of the tensor
[in]	virt_dim	virtual phase in each dimension
[in]	phase_rank	physical phase rank multiplied by virtual phase
[in]	vdata	array of input values
[out]	vpairs	pairs of keys and inputted values, allocated internally
[in,out]	nnz_blk	nonzero counts for each virtual block
[in]	sr	algstrct defining data type of array
[in]	edge_lda	lda of non-padded edge-lengths
[in]	f	sparsification filter

Definition at line 276 of file sparse_rw.cxx.

References alloc(), alloc_ptr(), ASSERT, cdealloc(), CTF_int::accumulatable::el_size, NS, CTF_int::algstrct::pair_alloc(), CTF_int::algstrct::pair_size(), CTF_int::algstrct::set_pair(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::tensor::sparsify().

void CTF_int::spspsum	(	algstrct const *	sr_A,
		int64_t	nA,
		ConstPairIterator	prs_A,
		char const *	beta,
		algstrct const *	sr_B,
		int64_t	nB,
		ConstPairIterator	prs_B,
		char const *	alpha,
		int64_t &	nnew,
		char *&	pprs_new,
		univar_function const *	func,
		int64_t	map_pfx
	)

As pairs in a sparse A set to the sparse set of elements defining the tensor, resulting in a set of size between nB and nB+nA.

Parameters

[in]	sr_A	algstrct defining data type of array
[in]	nA	number of elements in sparse tensor
[in]	prs_A	pairs of the sparse tensor
[in]	beta	scaling factor for data of the sparse tensor
[in]	sr_B	algstrct defining data type of array
[in]	nB	number of elements in the A set
[in]	prs_B	pairs of the A set
[in]	alpha	scaling factor for data of the A set
[out]	nnew	number of elements in resulting set
[out]	pprs_new	char array containing the pairs of the resulting set
[in]	func	NULL or pointer to a function to apply elementwise
[in]	map_pfx	how many times each element of A should be replicated

Definition at line 183 of file spr_seq_sum.cxx.

References ctf.core::a, CTF_int::univar_function::acc_f(), CTF_int::algstrct::add(), CTF_int::univar_function::apply_f(), ASSERT, ctf.core::b, CTF_int::algstrct::copy_pair(), CTF_int::accumulatable::el_size, CTF_int::univar_function::is_accumulator(), CTF_int::ConstPairIterator::k(), CTF_int::algstrct::mul(), CTF_int::algstrct::pair_alloc(), CTF_int::ConstPairIterator::read_val(), spA_spB_seq_sum(), TAU_FSTART, TAU_FSTOP, ctf.core::w, and CTF_int::PairIterator::write_val().

Referenced by spA_spB_seq_sum().

int CTF_int::stretch_virt	(	int	order,
		int	stretch_factor,
		mapping *	maps
	)

stretch virtualization by a factor

Parameters

[in]	order	number of maps to stretch
[in]	stretch_factor	factor to strech by
[in]	maps	mappings along each dimension to stretch

Definition at line 542 of file mapping.cxx.

References CTF_int::mapping::child, CTF_int::mapping::has_child, CTF_int::mapping::mapping(), CTF_int::mapping::np, PHYSICAL_MAP, SUCCESS, CTF_int::mapping::type, and VIRTUAL_MAP.

int CTF_int::strip_diag	(	int	order,
		int	order_tot,
		int const *	idx_map,
		int64_t	vrt_sz,
		mapping const *	edge_map,
		topology const *	topo,
		algstrct const *	sr,
		int *	blk_edge_len,
		int64_t *	blk_sz,
		strp_tsr **	stpr
	)

build stack required for stripping out diagonals of tensor

Parameters

[in]	order	number of dimensions of this tensor
[in]	order_tot	number of dimensions invovled in contraction/sum
[in]	idx_map	the index mapping for this contraction/sum
[in]	vrt_sz	size of virtual block
[in]	edge_map	mapping of each dimension
[in]	topo	topology the tensor is mapped to
[in]	sr	algstrct to be given to all stpr objs
[in,out]	blk_edge_len	edge lengths of local block after strip
[in,out]	blk_sz	size of local sub-block block after strip
[out]	stpr	class that recursively strips tensor

Returns: 1 if tensor needs to be stripped, 0 if not

Definition at line 273 of file strp_tsr.cxx.

References alloc_ptr(), CTF_int::strp_tsr::alloced, ASSERT, CTF_int::strp_tsr::blk_sz, CTF_int::strp_tsr::buffer, CTF_int::mapping::calc_phase(), CTF_int::mapping::calc_phys_phase(), CTF_int::mapping::calc_phys_rank(), cdealloc(), CTF_int::strp_tsr::edge_len, ctf.core::np(), CTF_int::strp_tsr::order, PHYSICAL_MAP, CTF_int::strp_tsr::sr_A, CTF_int::strp_tsr::strip_dim, CTF_int::strp_tsr::strip_idx, and VIRTUAL_MAP.

Referenced by CTF_int::summation::estimate_time(), CTF_int::scaling::execute(), and CTF_int::strp_ctr::strp_ctr().

void CTF_int::subsample	(	tensor *	A,
		double	probability
	)

extract a sample of the entries (if sparse of the current nonzeros)

Parameters

[in]	A	tensor to sample
[in]	probability	keep each entry with probability

Definition at line 103 of file ctf_ext.cxx.

References get_rand48(), IASSERT, CTF_int::tensor::sparsify(), and SUCCESS.

Referenced by ctf.core::MPI_Stop(), and ctf.core.tensor::sample().

int64_t CTF_int::sum_bool_tsr ( tensor * A )

sum all 1 values in boolean tensor

Parameters

[in] A tensor of boolean values

Returns: number of 1s in A

Definition at line 92 of file ctf_ext.cxx.

References ctf.core::a, CTF::Scalar< dtype >::get_val(), CTF_int::tensor::order, and CTF_int::tensor::wrld.

Referenced by ctf.core.tensor::__get__(), and ctf.core::MPI_Stop().

void CTF_int::sy_calc_idx_arr	(	int	order,
		int const *	lens,
		int const *	sym,
		int64_t	idx,
		int *	idx_arr
	)

same as above except assumes sym only NS or SY

Definition at line 121 of file util.cxx.

References ASSERT, ctf.core::dim, NS, and sy_packed_size().

Referenced by order_globally().

int64_t CTF_int::sy_packed_size	(	int	order,
		const int *	len,
		const int *	sym
	)

computes the size of a tensor in SY (NOT HOLLOW) packed symmetric layout

Parameters

[in]	order	tensor dimension
[in]	len	tensor edge _elngths
[in]	sym	tensor symmetries

Returns: size of tensor in packed layout

Definition at line 10 of file util.cxx.

Referenced by calc_cnt_displs(), compute_syoff(), cyclic_reshuffle(), CTF_int::seq_tsr_spctr::est_membw(), CTF_int::seq_tsr_ctr::est_membw(), CTF_int::tensor::est_time_unfold(), CTF_int::tensor::fold(), get_len_ordering(), glb_cyclic_reshuffle(), glb_ord_pup(), pad_cyclic_pup_virt_buff(), padded_reshuffle(), precompute_offsets(), CTF_int::tensor::set_padding(), spA_dnB_dnC_seq_ctr(), spA_dnB_seq_sum(), sy_calc_idx_arr(), sym_seq_ctr_cust(), sym_seq_ctr_ref(), sym_seq_sum_ref(), and CTF_int::tensor::unfold().

int CTF_int::sym_seq_ctr_cust	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		bivar_function const *	func
	)

performs symmetric contraction with custom elementwise function

Definition at line 566 of file sym_seq_ctr.cxx.

References CTF_int::bivar_function::acc_f(), CTF_int::algstrct::addid(), alloc(), ASSERT, cdealloc(), CHECK_SYM, compute_syoffs(), CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::isequal(), MAX_ORD, CTF_int::algstrct::mul(), CTF_int::algstrct::mulid(), RESET_IDX, CTF_int::algstrct::scal(), CTF_int::algstrct::set(), SWITCH_ORD_CALL, sy_packed_size(), sym_seq_ctr_loop(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_ctr::run().

int CTF_int::sym_seq_ctr_inr	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		iparam const *	prm,
		bivar_function const *	func
	)

performs symmetric contraction with blocked gemm

Definition at line 745 of file sym_seq_ctr.cxx.

References CTF_int::algstrct::addid(), alloc(), ASSERT, cdealloc(), CTF_int::bivar_function::cgemm(), CHECK_SYM, CTF_int::bivar_function::coffload_gemm(), CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, gemm(), CTF_int::algstrct::gemm_batch(), GET_MIN_MAX, inv_idx(), CTF_int::algstrct::isequal(), CTF_int::iparam::k, CTF_int::iparam::l, CTF_int::iparam::m, CTF_int::algstrct::mulid(), CTF_int::iparam::n, CTF_int::iparam::offload, CTF_int::algstrct::offload_gemm(), RESET_IDX, CTF_int::algstrct::scal(), CTF_int::algstrct::set(), CTF_int::iparam::sz_C, CTF_int::iparam::tA, TAU_FSTART, TAU_FSTOP, CTF_int::iparam::tB, and CTF_int::iparam::tC.

Referenced by CTF_int::seq_tsr_ctr::run().

template<int idim>

void CTF_int::sym_seq_ctr_loop	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 12 of file sym_seq_ctr.cxx.

References NS, and sym_seq_ctr_loop< 0 >().

Referenced by sym_seq_ctr_cust(), and sym_seq_ctr_ref().

template<>

void CTF_int::sym_seq_ctr_loop< 0 >	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 137 of file sym_seq_ctr.cxx.

References CTF_FLOPS_ADD, NS, and sym_seq_ctr_loop< MAX_ORD >().

Referenced by sym_seq_ctr_loop().

template void CTF_int::sym_seq_ctr_loop< MAX_ORD >	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C,
		uint64_t const	offsets_C,
		bivar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Referenced by sym_seq_ctr_loop< 0 >().

int CTF_int::sym_seq_ctr_ref	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		char const *	beta,
		char *	C,
		algstrct const *	sr_C,
		int	order_C,
		int const *	edge_len_C,
		int const *	sym_C,
		int const *	idx_map_C
	)

performs symmetric contraction with reference (unblocked) kernel

Definition at line 381 of file sym_seq_ctr.cxx.

References CTF_int::algstrct::add(), CTF_int::algstrct::addid(), alloc(), ASSERT, cdealloc(), CHECK_SYM, compute_syoffs(), CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::isequal(), MAX_ORD, CTF_int::algstrct::mul(), CTF_int::algstrct::mulid(), RESET_IDX, CTF_int::algstrct::scal(), CTF_int::algstrct::set(), SWITCH_ORD_CALL, sy_packed_size(), sym_seq_ctr_loop(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_ctr::run().

int CTF_int::sym_seq_scl_cust	(	char const *	alpha,
		char *	A,
		algstrct const *	sr_A,
		int const	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		endomorphism const *	func
	)

performs symmetric scaling using custom func

Definition at line 105 of file sym_seq_scl.cxx.

References alloc(), CTF_int::endomorphism::apply_f(), ASSERT, cdealloc(), CHECK_SYM, CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::mul(), RESET_IDX, sym_seq_sum_cust(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_scl::run(), and CTF_int::endomorphism::~endomorphism().

int CTF_int::sym_seq_scl_ref	(	char const *	alpha,
		char *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A
	)

performs symmetric scaling using algstrct const * sr_A

Definition at line 40 of file sym_seq_scl.cxx.

References alloc(), ASSERT, cdealloc(), CHECK_SYM, CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::mul(), RESET_IDX, sym_seq_sum_ref(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_scl::run(), and CTF_int::endomorphism::~endomorphism().

int CTF_int::sym_seq_sum_cust	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	beta,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		univar_function const *	func
	)

performs symmetric summation with custom elementwise function

Definition at line 524 of file sym_seq_sum.cxx.

References CTF_int::univar_function::acc_f(), alloc(), ASSERT, cdealloc(), CHECK_SYM, CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::mul(), RESET_IDX, SCAL_B, TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_sum::run(), and sym_seq_scl_cust().

int CTF_int::sym_seq_sum_inr	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	beta,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		int	inr_stride
	)

performs symmetric summation with blocked daxpy

Definition at line 434 of file sym_seq_sum.cxx.

References alloc(), ASSERT, CTF_int::algstrct::axpy(), cdealloc(), CHECK_SYM, CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), RESET_IDX, SCAL_B_inr, TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_sum::run().

template<int idim>

void CTF_int::sym_seq_sum_loop	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		univar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 11 of file sym_seq_sum.cxx.

References NS, and sym_seq_sum_loop< 0 >().

Referenced by sym_seq_sum_ref().

template<>

void CTF_int::sym_seq_sum_loop< 0 >	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		univar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Definition at line 107 of file sym_seq_sum.cxx.

References CTF_FLOPS_ADD, NS, and sym_seq_sum_loop< MAX_ORD >().

Referenced by sym_seq_sum_loop().

template void CTF_int::sym_seq_sum_loop< MAX_ORD >	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		uint64_t const	offsets_A,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B,
		uint64_t const	offsets_B,
		univar_function const *	func,
		int const *	idx,
		int const *	rev_idx_map,
		int	idx_max
	)

Referenced by sym_seq_sum_loop< 0 >().

int CTF_int::sym_seq_sum_ref	(	char const *	alpha,
		char const *	A,
		algstrct const *	sr_A,
		int	order_A,
		int const *	edge_len_A,
		int const *	sym_A,
		int const *	idx_map_A,
		char const *	beta,
		char *	B,
		algstrct const *	sr_B,
		int	order_B,
		int const *	edge_len_B,
		int const *	sym_B,
		int const *	idx_map_B
	)

performs symmetric contraction with unblocked reference kernel

Definition at line 292 of file sym_seq_sum.cxx.

References CTF_int::algstrct::add(), CTF_int::algstrct::addid(), alloc(), ASSERT, cdealloc(), CHECK_SYM, compute_syoffs(), CTF_FLOPS_ADD, CTF_int::accumulatable::el_size, GET_MIN_MAX, inv_idx(), CTF_int::algstrct::isequal(), MAX_ORD, CTF_int::algstrct::mul(), CTF_int::algstrct::mulid(), RESET_IDX, CTF_int::algstrct::scal(), SCAL_B, CTF_int::algstrct::set(), SWITCH_ORD_CALL, sy_packed_size(), sym_seq_sum_loop(), TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::seq_tsr_sum::run(), and sym_seq_scl_ref().

void CTF_int::symmetrize	(	tensor *	sym_tsr,
		tensor *	nonsym_tsr
	)

folds the data of a tensor

Parameters

[in]	sym_tsr	starting symmetric tensor (where data ends)
[in]	nonsym_tsr	new tensor with a potentially unfolded symmetry

Definition at line 237 of file symmetrization.cxx.

References CTF_int::algstrct::addid(), CTF_int::algstrct::addinv(), alloc(), alloc_ptr(), AS, CTF_int::tensor::calc_nvirt(), CTF_int::algstrct::cast_double(), cdealloc(), CTF_int::algstrct::copy(), copy_mapping(), CTF_int::tensor::data, CTF_int::tensor::edge_map, CTF_int::accumulatable::el_size, CTF_int::scaling::execute(), CTF_int::summation::execute(), CTF_int::tensor::is_home, CTF_int::tensor::is_mapped, CTF_int::tensor::is_sparse, MAX, CTF_int::algstrct::mulid(), CTF_int::tensor::name, CTF_int::tensor::nnz_blk, NS, CTF_int::tensor::order, CTF_int::tensor::profile, CTF::World::rank, scal_diag(), CTF_int::tensor::set_new_nnz_glb(), CTF_int::tensor::set_padding(), SH, CTF_int::tensor::size, CTF_int::tensor::sr, CTF::Timer::start(), CTF::Timer::stop(), CTF_int::summation::sum_tensors(), SY, CTF_int::tensor::sym, TAU_FSTART, TAU_FSTOP, CTF_int::tensor::topo, VPRINTF, CTF_int::tensor::wrld, and CTF_int::tensor::zero_out_padding().

Referenced by CTF_int::summation::estimate_time(), and get_len_ordering().

template void CTF_int::tensor::compare_elementwise< std::complex< double > >	(	tensor *	A,
		tensor *	B
	)

template void CTF_int::tensor::compare_elementwise< std::complex< float > >	(	tensor *	A,
		tensor *	B
	)

bool CTF_int::try_mkl_coo_to_csr	(	int64_t	nz,
		int	nrow,
		char *	csr_vs,
		int *	csr_ja,
		int *	csr_ia,
		char const *	coo_vs,
		int const *	coo_rs,
		int const *	coo_cs,
		int	el_size
	)

Definition at line 70 of file set.cxx.

Referenced by seq_coo_to_csr().

bool CTF_int::try_mkl_csr_to_coo	(	int64_t	nz,
		int	nrow,
		char const *	csr_vs,
		int const *	csr_ja,
		int const *	csr_ia,
		char *	coo_vs,
		int *	coo_rs,
		int *	coo_cs,
		int	el_size
	)

Definition at line 91 of file set.cxx.

Referenced by seq_csr_to_coo().

int CTF_int::untag_mem ( void * ptr )

stops tracking memory allocated by CTF, so user doesn't have to call free

Parameters

[in,out] ptr pointer to set to address to free

Definition at line 376 of file memcontrol.cxx.

References ABORT, ERROR, ctf.core::it, and SUCCESS.

void CTF_int::update_all_models ( MPI_Comm comm )

Definition at line 15 of file model.cxx.

References get_all_models().

Referenced by CTF_int::Model::dump_data(), sparse_mp3(), and train_all().

void CTF_int::wr_pairs_layout	(	int	order,
		int	np,
		int64_t	inwrite,
		char const *	alpha,
		char const *	beta,
		char	rw,
		int	num_virt,
		int const *	sym,
		int const *	edge_len,
		int const *	padding,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_phase,
		int *	virt_phys_rank,
		int const *	bucket_lda,
		char *	wr_pairs_buf,
		char *	rw_data,
		CommData	glb_comm,
		algstrct const *	sr,
		bool	is_sparse,
		int64_t	nnz_loc,
		int64_t *	nnz_blk,
		char *&	pprs_new,
		int64_t &	nnz_loc_new
	)

read or write pairs from / to tensor

Parameters

[in]	order	tensor dimension
[in]	np	number of processors
[in]	inwrite	number of pairs
[in]	alpha	multiplier for new value
[in]	beta	multiplier for old value
[in]	rw	whether to read 'r' or write 'w'
[in]	num_virt	new total virtualization factor
[in]	sym	symmetries of tensor
[in]	edge_len	tensor edge lengths
[in]	padding	padding of tensor
[in]	phase	total phase in each dimension
[in]	phys_phase	physical phase in each dimension
[in]	virt_phase	virtualization in each dimension
[in]	virt_phys_rank	virtualized rank in total phase
[in]	bucket_lda	prefix sum of the processor grid
[in,out]	wr_pairs_buf	pairs to read or write
[in,out]	rw_data	data to read from or write to
[in]	glb_comm	the global communicator
[in]	sr	algstrct context defining values
[in]	is_sparse	if true local data (vdata) is sparse, otherwise rest of params irrelevant
[in]	nnz_loc	number of local elements in sparse tensor
[in]	nnz_blk	number of local elements in each block
[out]	pprs_new	new elements of the local sparse tensor (if rw=='w')
[out]	nnz_loc_new	number of elements of the new local sparse tensor (if rw=='w')

Definition at line 872 of file sparse_rw.cxx.

References CTF_int::algstrct::addid(), CTF_int::algstrct::addinv(), CTF_int::CommData::all_to_allv(), alloc_ptr(), AS, ASSERT, bucket_by_pe(), bucket_by_virt(), cdealloc(), CTF_int::CommData::cm, cvrt_idx(), DEBUG_PRINTF, CTF_int::accumulatable::el_size, CTF_int::PairIterator::k(), CTF_int::PairIterator::lower_bound(), CTF_int::CommData::np, ctf.core::np(), NS, pad_key(), CTF_int::algstrct::pair_alloc(), CTF_int::algstrct::pair_dealloc(), CTF_int::algstrct::pair_size(), CTF_int::PairIterator::ptr, CTF_int::CommData::rank, CTF_int::PairIterator::read(), readwrite(), SH, sign(), CTF_int::PairIterator::sort(), sp_read(), sp_write(), TAU_FSTART, TAU_FSTOP, CTF_int::PairIterator::write_key(), and CTF_int::PairIterator::write_val().

Referenced by padded_reshuffle(), and CTF_int::tensor::write().

void CTF_int::write_all_models ( std::string file_name )

Definition at line 42 of file model.cxx.

References get_all_models().

Referenced by CTF_int::Model::dump_data(), and train_all().

template<typename dtype >

void CTF_int::write_data_mpiio	(	CTF::World const *	dw,
		char const *	fpath,
		char *	datastr,
		int64_t	str_len
	)

write sparse tensor data to file using MPI-I/O, from string with one entry per line (different entries on each process)

Parameters

[in]	dw	world (comm)
[in]	fpath	file name
[in]	datastr	array of strings to write to file
[in]	str_len	num chars in string

Definition at line 214 of file graph_io_aux.cxx.

References CTF::World::comm.

void CTF_int::zero_padding	(	int	order,
		int64_t	size,
		int	nvirt,
		int const *	edge_len,
		int const *	sym,
		int const *	padding,
		int const *	phase,
		int const *	phys_phase,
		int const *	virt_phase,
		int const *	cphase_rank,
		char *	vdata,
		algstrct const *	sr
	)

sets to zero all values in padded region of tensor

brief pads a tensor param[in] order tensor dimension param[in] num_pair number of pairs param[in] edge_len tensor edge lengths param[in] sym symmetries of tensor param[in] padding padding of tensor (included in edge_len) param[in] phys_phase phase of the tensor on virtualized processor grid param[in] phase_rank physical phase rank multiplied by virtual phase param[in] virt_phase virtual phase in each dimension param[in] old_data array of input pairs param[out] new_pairs padded pairs param[out] new_size number of new padded pairs param[in] algstrct defines sizeo of each pair

Parameters

[in]	order	tensor dimension
[in]	size	number of values
[in]	nvirt	total virtualization factor
[in]	edge_len	tensor edge lengths with padding
[in]	sym	symmetries of tensor
[in]	padding	how much of the edge lengths is padding
[in]	phase	phase of the tensor on virtualized processor grid
[in]	phys_phase	phase of the tensor on virtualized processor grid
[in]	virt_phase	virtual phase in each dimension
[in]	cphase_rank	physical phase rank
[in,out]	vdata	array of all local data
[in]	sr	algstrct defines sizeo of each pair

Definition at line 374 of file pad.cxx.

References CTF_int::algstrct::addid(), alloc_ptr(), AS, ASSERT, calc_idx_arr(), cdealloc(), ctf.core::dim, CTF_int::accumulatable::el_size, MIN, NS, CTF_int::algstrct::set(), SH, SY, TAU_FSTART, and TAU_FSTOP.

Referenced by CTF_int::tensor::zero_out_padding().

Variable Documentation

LinModel<3> CTF_int::allred_mdl(allred_mdl_init,"allred_mdl")

Referenced by get_rand48().

LinModel<3> CTF_int::allred_mdl_cst(allred_mdl_cst_init,"allred_mdl_cst")

Referenced by get_rand48().

double CTF_int::allred_mdl_cst_init = {-3.3754E-04, 2.1343E-04, 3.0801E-09}

Definition at line 35 of file init_models.cxx.

Referenced by get_rand48().

double CTF_int::allred_mdl_init = {8.4416E-07, 6.8651E-06, 3.5845E-08}

Definition at line 34 of file init_models.cxx.

Referenced by get_rand48().

LinModel<3> CTF_int::alltoall_mdl(alltoall_mdl_init,"alltoall_mdl")

Referenced by get_rand48().

double CTF_int::alltoall_mdl_init = {1.0000E-06, 1.0000E-06, 5.0000E-10}

Definition at line 30 of file init_models.cxx.

Referenced by get_rand48().

LinModel<3> CTF_int::alltoallv_mdl(alltoallv_mdl_init,"alltoallv_mdl")

Referenced by get_rand48().

double CTF_int::alltoallv_mdl_init = {2.7437E-06, 2.2416E-05, 1.0469E-08}

Definition at line 31 of file init_models.cxx.

Referenced by get_rand48().

LinModel<3> CTF_int::bcast_mdl(bcast_mdl_init,"bcast_mdl")

Referenced by get_rand48().

double CTF_int::bcast_mdl_init = {1.5045E-06, 1.4485E-05, 3.2876E-09}

Definition at line 36 of file init_models.cxx.

Referenced by get_rand48().

LinModel<2> CTF_int::blres_mdl(blres_mdl_init,"blres_mdl")

Referenced by calc_cnt_displs().

double CTF_int::blres_mdl_init = {1.0598E-05, 7.2741E-08}

Definition at line 29 of file init_models.cxx.

Referenced by calc_cnt_displs().

char* CTF_int::cpy_buffer[CPY_BUFFER_SIZE]

Definition at line 105 of file memcontrol.cxx.

LinModel<3> CTF_int::csrred_mdl(csrred_mdl_init,"csrred_mdl")

Referenced by CTF_int::algstrct::csr_reduce(), and CTF_int::algstrct::estimate_csr_red_time().

LinModel<3> CTF_int::csrred_mdl_cst(csrred_mdl_cst_init,"csrred_mdl_cst")

double CTF_int::csrred_mdl_cst_init = {-1.8323E-04, 1.3076E-04, 2.8732E-09}

Definition at line 39 of file init_models.cxx.

double CTF_int::csrred_mdl_init = {3.7005E-05, 1.1854E-04, 5.5165E-09}

Definition at line 38 of file init_models.cxx.

LinModel<3> CTF_int::dgtog_res_mdl(dgtog_res_mdl_init,"dgtog_res_mdl")

Referenced by dgtog_est_time(), and dgtog_reshuffle().

double CTF_int::dgtog_res_mdl_init = {2.9786E-05, 2.4335E-04, 1.0845E-08}

Definition at line 28 of file init_models.cxx.

CTF::Ring<double> CTF_int::double_ring = CTF::Ring<double>()

Definition at line 9 of file ring.cxx.

Referenced by get_double_ring().

double CTF_int::download_mdl_init[]

CTF::Ring<float> CTF_int::float_ring = CTF::Ring<float>()

Definition at line 5 of file ring.cxx.

Referenced by get_float_ring().

int CTF_int::instance_counter = 0

Definition at line 75 of file memcontrol.cxx.

Referenced by get_num_instances().

CTF::Ring<int64_t> CTF_int::int64_t_ring = CTF::Ring<int64_t>()

Definition at line 17 of file ring.cxx.

Referenced by get_int64_t_ring().

CTF::Ring<int> CTF_int::int_ring = CTF::Ring<int>()

Definition at line 13 of file ring.cxx.

Referenced by get_int_ring().

LinModel<2> CTF_int::long_contig_transp_mdl(long_contig_transp_mdl_init,"long_contig_transp_mdl")

Referenced by est_time_transp(), and nosym_transpose().

double CTF_int::long_contig_transp_mdl_init = {2.9158E-10, 3.0501E-09}

Definition at line 25 of file init_models.cxx.

int CTF_int::max_threads

Definition at line 74 of file memcontrol.cxx.

Referenced by cdealloc(), cdealloc_cond(), mem_create(), and mem_exit().

int64_t CTF_int::mem_size = 0

Definition at line 72 of file memcontrol.cxx.

Referenced by CTF::World::~World().

std::list<mem_loc> CTF_int::mem_stacks[MAX_THREADS]

Definition at line 90 of file memcontrol.cxx.

int64_t CTF_int::mem_used[MAX_THREADS]

Definition at line 76 of file memcontrol.cxx.

double CTF_int::memcap = 0.5

Definition at line 71 of file memcontrol.cxx.

Referenced by proc_bytes_available().

MPI_Datatype CTF_int::MPI_CTF_BOOL = MPI_CXX_BOOL

Definition at line 14 of file set.cxx.

Referenced by get_default_mdtype(), and get_default_mdtype< bool >().

MPI_Datatype CTF_int::MPI_CTF_DOUBLE_COMPLEX = MPI_CXX_DOUBLE_COMPLEX

Definition at line 15 of file set.cxx.

Referenced by cvrt_idx(), get_default_mdtype(), get_default_mdtype< std::complex< double > >(), and get_mpi_dt().

MPI_Datatype CTF_int::MPI_CTF_LONG_DOUBLE_COMPLEX = MPI_CXX_LONG_DOUBLE_COMPLEX

Definition at line 16 of file set.cxx.

Referenced by get_default_mdtype(), and get_default_mdtype< std::complex< long double > >().

std::list<mem_loc> CTF_int::mst

Definition at line 102 of file memcontrol.cxx.

Referenced by contract_mst(), and get_mst().

void* CTF_int::mst_buffer = 0

Definition at line 94 of file memcontrol.cxx.

int64_t CTF_int::mst_buffer_ptr = 0

Definition at line 100 of file memcontrol.cxx.

Referenced by mem_exit(), and mst_alloc_ptr().

int64_t CTF_int::mst_buffer_size = 0

Definition at line 96 of file memcontrol.cxx.

int64_t CTF_int::mst_buffer_used = 0

Definition at line 98 of file memcontrol.cxx.

Referenced by mst_alloc_ptr().

LinModel<2> CTF_int::non_contig_transp_mdl(non_contig_transp_mdl_init,"non_contig_transp_mdl")

Referenced by est_time_transp(), and nosym_transpose().

double CTF_int::non_contig_transp_mdl_init = {4.0475E-08, 4.0463E-09}

Definition at line 27 of file init_models.cxx.

LinModel<2> CTF_int::pin_keys_mdl(pin_keys_mdl_init,"pin_keys_mdl")

Referenced by CTF_int::spctr_pin_keys::spctr_pin_keys().

double CTF_int::pin_keys_mdl_init = {3.1189E-09, 6.6717E-08}

Definition at line 18 of file init_models.cxx.

Referenced by CTF_int::spctr_pin_keys::spctr_pin_keys().

LinModel<3> CTF_int::red_mdl(red_mdl_init,"red_mdl")

Referenced by get_rand48().

LinModel<3> CTF_int::red_mdl_cst(red_mdl_cst_init,"red_mdl_cst")

Referenced by get_rand48().

double CTF_int::red_mdl_cst_init = {5.7302E-07, 4.7347E-06, 6.0191E-10}

Definition at line 33 of file init_models.cxx.

Referenced by get_rand48().

double CTF_int::red_mdl_init = {6.2935E-07, 4.6276E-06, 9.2245E-10}

Definition at line 32 of file init_models.cxx.

Referenced by get_rand48().

std::mersenne_twister_engine<std::uint_fast64_t, 64, 312, 156, 31, 0xb5026f5aa96619e9, 29, 0x5555555555555555, 17, 0x71d67fffeda60000, 37, 0xfff7eee000000000, 43, 6364136223846793005> CTF_int::rng

Definition at line 20 of file common.cxx.

Referenced by get_rand48(), and init_rng().

std::set< grid_wrapper > CTF_int::scalapack_grids

index for ScaLAPACK processor grids

Definition at line 27 of file world.cxx.

Referenced by CTF::Matrix< dtype >::get_desc(), and CTF::World::~World().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_cst(seq_tsr_ctr_mdl_cst_init,"seq_tsr_ctr_mdl_cst")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_cst_init = {5.1626E-06, -6.3215E-11, 3.9638E-09}

Definition at line 19 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_cst_inr(seq_tsr_ctr_mdl_cst_inr_init,"seq_tsr_ctr_mdl_cst_inr")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_cst_inr_init = {1.3863E-04, 2.0119E-10, 9.8820E-09}

Definition at line 23 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_cst_off(seq_tsr_ctr_mdl_cst_off_init,"seq_tsr_ctr_mdl_cst_off")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_cst_off_init = {8.4844E-04, -5.9246E-11, 3.5247E-10}

Definition at line 24 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_inr(seq_tsr_ctr_mdl_inr_init,"seq_tsr_ctr_mdl_inr")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_inr_init = {2.0647E-08, 1.9721E-10, 2.9948E-11}

Definition at line 21 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_off(seq_tsr_ctr_mdl_off_init,"seq_tsr_ctr_mdl_off")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_off_init = {6.2925E-05, 1.7449E-11, 1.7211E-12}

Definition at line 22 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_ctr_mdl_ref(seq_tsr_ctr_mdl_ref_init,"seq_tsr_ctr_mdl_ref")

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

double CTF_int::seq_tsr_ctr_mdl_ref_init = {4.9138E-08, 5.8290E-10, 4.8575E-11}

Definition at line 20 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_ctr::mem_fp().

LinModel<3> CTF_int::seq_tsr_spctr_cst_k0(seq_tsr_spctr_cst_k0_init,"seq_tsr_spctr_cst_k0")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_k0_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 8 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_k1(seq_tsr_spctr_cst_k1_init,"seq_tsr_spctr_cst_k1")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_k1_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 9 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_k2(seq_tsr_spctr_cst_k2_init,"seq_tsr_spctr_cst_k2")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_k2_init = {-8.8459E-08, 8.1207E-10, -2.8486E-12}

Definition at line 10 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_k3(seq_tsr_spctr_cst_k3_init,"seq_tsr_spctr_cst_k3")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_k3_init = {1.8504E-08, 2.9154E-11, 2.1973E-11}

Definition at line 11 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_k4(seq_tsr_spctr_cst_k4_init,"seq_tsr_spctr_cst_k4")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_k4_init = {2.0948E-05, 1.2294E-09, 8.0037E-10}

Definition at line 12 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_off_k0(seq_tsr_spctr_cst_off_k0_init,"seq_tsr_spctr_cst_off_k0")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_off_k0_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 2 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_off_k1(seq_tsr_spctr_cst_off_k1_init,"seq_tsr_spctr_cst_off_k1")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_off_k1_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 3 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_cst_off_k2(seq_tsr_spctr_cst_off_k2_init,"seq_tsr_spctr_cst_off_k2")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_cst_off_k2_init = {-2.1996E-04, 3.1883E-09, 3.8743E-11}

Definition at line 4 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_k0(seq_tsr_spctr_k0_init,"seq_tsr_spctr_k0")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_k0_init = {2.2620E-08, -5.7494E-10, 2.2146E-09}

Definition at line 13 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_k1(seq_tsr_spctr_k1_init,"seq_tsr_spctr_k1")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_k1_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 14 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_k2(seq_tsr_spctr_k2_init,"seq_tsr_spctr_k2")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_k2_init = {3.0917E-08, 5.2181E-11, 4.1634E-12}

Definition at line 15 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_k3(seq_tsr_spctr_k3_init,"seq_tsr_spctr_k3")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_k3_init = {7.2456E-08, 1.5128E-10, -1.5528E-12}

Definition at line 16 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_k4(seq_tsr_spctr_k4_init,"seq_tsr_spctr_k4")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_k4_init = {1.6880E-07, 4.9411E-10, 9.2847E-13}

Definition at line 17 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_off_k0(seq_tsr_spctr_off_k0_init,"seq_tsr_spctr_off_k0")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_off_k0_init = {8.6970E-06, 4.5598E-11, 1.1544E-09}

Definition at line 5 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_off_k1(seq_tsr_spctr_off_k1_init,"seq_tsr_spctr_off_k1")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_off_k1_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 6 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<3> CTF_int::seq_tsr_spctr_off_k2(seq_tsr_spctr_off_k2_init,"seq_tsr_spctr_off_k2")

Referenced by CTF_int::seq_tsr_spctr::est_membw().

double CTF_int::seq_tsr_spctr_off_k2_init = {5.3745E-06, 3.6464E-08, 2.2334E-10}

Definition at line 7 of file init_models.cxx.

Referenced by CTF_int::seq_tsr_spctr::est_membw().

LinModel<2> CTF_int::shrt_contig_transp_mdl(shrt_contig_transp_mdl_init,"shrt_contig_transp_mdl")

Referenced by est_time_transp(), and nosym_transpose().

double CTF_int::shrt_contig_transp_mdl_init = {1.3427E-08, 4.3168E-09}

Definition at line 26 of file init_models.cxx.

LinModel<3> CTF_int::spredist_mdl(spredist_mdl_init,"spredist_mdl")

Referenced by CTF_int::tensor::redistribute(), and spredist_est_time().

double CTF_int::spredist_mdl_init = {1.2744E-04, 1.0278E-03, 7.6837E-08}

Definition at line 37 of file init_models.cxx.

int64_t CTF_int::tot_mem_available = -1

Definition at line 78 of file memcontrol.cxx.

Referenced by proc_bytes_total().

int64_t CTF_int::tot_mem_used

Definition at line 77 of file memcontrol.cxx.

Referenced by main(), and proc_bytes_used().

int64_t CTF_int::total_flop_count = 0

Definition at line 171 of file common.cxx.

Referenced by get_flops().

double CTF_int::upload_mdl_init[]

Data Structures

Typedefs

Enumerations

Functions

Variables

Typedef Documentation

Enumeration Type Documentation

Function Documentation

Variable Documentation